About Cleveland ASA --- Regular Meetings --- Conferences--- Workshops
K-12 Programs --- Officers --- Membership Info ---Links --- Chapter News
Current Monthly Meeting

Cleveland Chapter of the ASA

Past Meetings

2007 2006 2005 2004 2003 2002 2001 2000
1999 1998 1997 1996 1995 1994 1993 1992 1991 1990
1989 1986 1985 1984 1983 1982 1981 1980
1974 1971 1970
1968 1966 1965
1952


2007 Talks
Wednesday, December 12, 2007
Building Risk-Adjustment Models for the Assessment of Obstetrical Quality :
Jennifer L. Bailit and Thomas E. Love

In this talk we will discuss the data and statistical challenges involved in creating risk-adjustment models for the evaluation of obstetrical care.

Many traditional measures of obstetrical quality, such as raw hospital cesarean rates, are becoming obsolete. When cesarean rates are compared without risk adjustment for maternal risk factors, hospitals treating complex patients appear to be providing poor quality care when in fact they may be providing superb care. Additionally, hospitals providing care to low risk patients may appear to be providing high quality care when in fact they are not.
The statistical and other methodological considerations related to producing valuable risk-adjustment measures in this context are substantial, especially in light of the many additional issues related to working in a "hot button" field like the assessment of obstetrical care. We will share some of our more interesting experiences from the varied perspectives of a clinical researcher and a biostatistician.

Friday, March 30, 2007
Adventures in Teaching with Technology :
Webster West

There are a tremendous number of technology resources available for teaching statistics. These range from interactive web-based applets to video lectures. The trick is often times determining which resources actually work and how best to incorporate them into your courses. Techniques for evaluating technology learning tools and some common integration approaches will be discussed. A discussion of the potential impact of technology on the future of statistics education will also be provided.


2006 Talks
Wednesday, November 1, 2006
Missing Data Methodology in Malaria Studies :
Rhoderick Machekano

Efficacy studies of malaria treatments can be plagued by indeterminate outcomes for some patients. The study motivating this work defines the outcome of interest (treatment failure) as recrudescence and for some subjects, it is unclear whether a recurrence of malaria is due to that or new infection. This results in a specific kind of missing data. The effect of missing data in causal inference problems is widely recognized. Methods that adjust for possible bias from missing data include a variety of imputation procedures (extreme case analysis, hot-deck, single and multiple imputation), inverse weighting methods, and likelihood based methods (data augmentation, EM procedures and their extensions). In this talk, I focus on multiple imputation, two inverse weighting procedures (the inverse probability weighted (IPW) and the doubly robust (DR) estimators), and a likelihood based methodology (Gcomputation), comparing the methods' applicability to the efficient estimation of malaria treatments effects. I present results from simulation studies as well as results from an application to malaria efficacy studies from Uganda.

Wednesday, September 20, 2006
What Mathematics and Forrest Gump Teach Us About Lotteries :
Ron Wasserstein

Since the inception of the Kansas Lottery in 1987, I have been speaking to school and civic groups about the lottery--how it works, what the probability of winning is and how that is computed, and most importantly, what the probability is of coming out ahead (a winner!) in the lottery. Initial efforts involved a standard lecture, but before long it became clear that more was needed. Thus, I have developed a computer game that simulates playing the lottery multiple times. Playing the game is not only fun, but it also shows students vividly what happens in the long run. Accompanying the computer game is a set of PowerPoint slides which teach about the lottery and probability. In this presentation to the Cleveland Chapter, I will "teach" this subject, demonstrating how these tools can be used as outreach to local groups. I will also provide the computer game and slides on CD for free use by chapter members.

Wednesday, April 12, 2006
Bayesian and Frequentist Methods for Provider Profiling Using Risk-Adjusted Assessments of Medical Outcomes :
Joe Sedransk

We propose a new method and compare conventional and Bayesian methodologies that are used or proposed for use for 'provider profiling,' an evaluation of the quality of health care. The conventional approaches to computing these provider assessments are to use likelihood-based frequentist methodologies, and the new Bayesian method is patterned after these. For each of three models we compare the frequentist and Bayesian approaches using the data employed by the New York State Department of Health for its annually released reports that profile hospitals permitted to perform coronary artery bypass graft surgery. Additional, constructed, data sets are used to sharpen our conclusions. With the advances of Markov Chain Monte Carlo methods, Bayesian methods are easily implemented and are preferable to standard frequentist methods for models with a binary dependent variable since the latter always rely on asymptotic approximations.
Comparisons across methods associated with different models are important because of current proposals to use random effect (exchangeable) models for provider profiling. We also summarize and discuss important issues in the conduct of provider profiling such as inclusion of provider characteristics in the model and choice of criteria for determining unsatisfactory performance.


2005 Talks
December 15, 2005
Hands-on Bayesian Data Analysis using WinBUGS [doubled as fall workshop] :
William F. Guthrie, National Institute of Standards and Technology

This workshop is designed to provide statisticians, scientists, or engineers with the tools necessary to begin to use Bayesian inference in applied problems. Participants in the course will learn the basics of Bayesian modeling and inference using Markov chain Monte Carlo simulation with the open-source software package WinBUGS.

The workshop will introduce some of the theory underlying Bayesian analysis, but will primarily focus on Bayesian analysis of "real-world" scientific applications using examples from collaborative research with NIST scientists and engineers. Topics discussed will include Bayesian modeling, Markov chain Monte Carlo algorithms, convergence tests, model validation, and inference.

September 14, 2005
SAT scores for sale? Assessment of Commercial Test Preparation Via Optimal Full Matching. :
Ben B. Hansen, University of Michigan

Post-stratification is an old, flexible, efficient, and conceptually plain statistical technique. If a treatment and a control group are to be compared, and if every treated subject is sufficiently similar to one or more controls as to justify comparison to it, then with the right stratification one can rightly estimate treatment effects simply by averaging and differencing outcomes.

An impediment is that post-stratification is practically feasible, usually, only when there are few covariates. Often there are many covariates, not just one or two, on which subjects should be similar so as to justify their comparison. For the purpose of estimating treatment effects, this issue is dispensed with by propensity scores, which reduce the dimension of the covariates to one. Because of this, observational studies are increasingly analyzed by way of stratification along a propensity score. Commonly used methods for stratifying along an estimated propensity include pair matching, matching with multiple controls, and subclassification along quintiles of the score.

Now the "right" stratification --- one that pairs only subjects with sufficiently similar values of the estimated score --- need not take a simple form, in which case the commonly used methods of stratification carry no guarantee of finding it. However, there is always a so-called full matching that is at least as good as any other stratification. This will be illustrated with a case study of full matching, with propensity scores, applied to estimate effects of commercial coaching for the SAT. I have created an add-on package to R to perform optimal full matching, "optmatch," and I shall also illustrate its use.
June 1, 2005
When Statistical Process Control Outweighs Randomized Clinical Trials :
Dr. Mireya Diaz-Insua, Case Western Reserve University

Clinical trials are the pinnacle in the hierarchy of evidence-based medicine. The main reason being the control they exert over observable and non-observable bias by virtue of the randomization process. We find proof of the use of trials as far back as verses in the book of Daniel in the Old Testament. In the late 1920s, Amberson designed the first randomized trial in medicine (known to us) to assess the efficacy of sanocrysin in the treatment of tuberculosis, 43 years after Peirce incorporated randomization in experimentation. Despite its established validity in the chain of evidence, there are circumstances when a randomized trial is not feasible because of time constraints and/or ethical concerns. In these situations, viable alternatives are necessary. Vaccine or treatment for diminishing the effect of outbreaks, parachute jumping, lethal conditions that require approaches with fast results are just examples of such scenarios. We show here tools from Statistical Process Control (SPC) that present a feasible alternative for those cases. We show the implementation of such strategy within the context of a specific case study of vaccine development and testing for rabies conducted by Pasteur in the 1880s, when randomization in experimentation was almost unknown. SPC tools will prove through this example how and when they are a plausible alternative to the paradigmatic randomized clinical trial. This is joint work with Dr. Duncan Neuhauser, and together they have produced three recent articles on related topics for the journal Quality and Safety in Health Care.

2004 Talks
December 1, 2004
Calls for 911 Service: A Collaboration Between CSU and the Cleveland City Police :
Dr. John P. Holcomb, Jr., Cleveland State University

In the last two years, professors from the Cleveland State University Department of Mathematics and Sociology have been working closely with the City of Cleveland Police Department. This partnership has resulted in access to police records cataloging all 911 calls for the city since 1995. In the presentation, I will share summary graphs and statistics as it relates to calls for service throughout the city and by various districts. I will also share various models that undergraduate and graduate students have developed to predict the number of calls for service.
September 1, 2004
Regression Modeling of Left and Right-Censored Outcomes :
Jeff Hammel, Cleveland Clinic

This presentation is motivated by the study of the susceptibility of bacteria to antimicrobial drugs. Susceptibility (or its opposite, resistance) is often measured by the minimum drug concentration required to kill the bacteria, or simply the minimum inhibitory concentration (MIC).

An MIC can be measured for a bacteria sample from a patient diagnosed with an infection. The MIC is often determined in a laboratory using a finite set of titred drug concentrations. When the MIC is below the lowest tested concentration or above the largest tested concentration, then the recorded value is left- or right-censored, respectively. That is, it is of the form '≤ k1' or '>k2'. The tested concentrations are also typically integer powers of 2 mg/L, in which case the censored values can be written as '≤2L' or '>2R' for integer values of L and R.

MIC data of this form is easily analyzed using either SAS or S-plus. I will demonstrate code for doing these analyses.
June 2, 2004
An Introduction to the R Language :
Ethan Katz, Cleveland Clinic
Jason Connor, Carnegie Mellon University
A basic, interactive, hands-on introduction to the R language, including statistical analysis and graphics procedures. Topics include:

Getting started with R: downloading R, obtaining documentation, and using the help system.
Statistics with R: descriptive statistics, statistical tests, and model fitting.
Graphics with R: basics and examples
February 27, 2004
Analytical Tools in the SAS System :
William Kuhfield, SAS Institute
Andrew Karp, Sierra Information Services, Inc.
Dr. Kuhfield, (a Cleveland-area native) who is a leading developer in the Research and Development group at SAS Worldwide Headquarters in Cary, NC., will present an introduction to the field of experimental design. His talk will focus on an easy-to-use tool for building efficient experimental designs.
The second presentation is a tutorial on how to build and interpret predictive models using PROC LOGISTIC. This analytic technique is frequently used in both medical/scientific and business analytic projects to predict the probability that an event (e.g., product purchase, disease outcome) will occur. This tutorial will be given by Andrew Karp, a well-known SAS Software consultant, trainer, and SAS user group meeting presenter.

2003 Talks
December 3, 2003
Propensity Scores: What Do They Do, How Should I Use Them, And Why Should I Care? :
Thomas E. Love,
Case Western Reserve University and MetroHealth Medical Center
Many statistical problems aim to draw causal inferences about the effects of policies, treatments or interventions. But the data in most cases are observational, rather than experimental - that is, the data are collected through the observation of systems as they operate in normal practice, rather than under carefully controlled conditions. In particular, the investigators have no control over the assignments of exposures. Such data are relatively inexpensive to obtain, and may represent the "real world" more effectively than the results of randomized experiments. However, in using an observational study to estimate a treatment effect, the "exposed" and "control" groups often differ substantially in terms of background characteristics. This talk will provide a friendly introduction to propensity score methods for reducing the impact of selection bias on observational studies.
October 27, 2003 (Fall Workshop)
Analysis of Curves: Esteban Walker, University of Tennessee
Advances in technology have dramatically increased the amount and quality of data that are recorded in all areas of human endeavor. Thousands of measurements are available nowadays in situations where previously only a few measurements, at given points in time or space, were taken. These measurements allow the reconstruction of the whole profile or "signature". Basically, the profile becomes the unit of analysis. This seminar will discuss two problems with profiles: (1) how to determine if predetermined sets of curves are different, and (2) how to identify clusters in a set of curves. Examples from various fields will be presented. Instructions on the implementation of these techniques in S-Plus and SAS will be given.
September 17, 2003
Reversible Jump Markov Chain Monte Carlo For Linear Regression Models: Patrick Gaffney, Lubrizol
Reversible Jump Markov Chain Monte Carlo (RJ-MCMC) is increasing in popularity as witnessed by the number of citations to the seminal paper by Green. This methodology gives the statistician a means of solving difficult problems where a parametric model has been proposed but the number of terms in the model is unknown. Some examples include change-point analysis, image analysis, and gene discovery. Several issues arise in practice: initialization of the chain, the proposal mechanism for parameters, correct formation of the acceptance ratio, and final model selection. In this talk, I will address these issues when RJ-MCMC is applied to solve linear regression models in a Bayesian framework. The lessons learned can readily be applied to more complex situations.
June 4, 2003
An Introduction to Value-Added Methods for Teacher Accountability: J.R. Lockwood, RAND Statistics Group
As underscored by the federal No Child Left Behind Act of 2001, a currently active and central education policy initiative is the use of scores on standardized achievement tests to hold educators accountable for student outcomes. One of the challenges to this endeavor is combining test score information into a single measure that provides evidence of school or teacher effectiveness. Although many states and districts rely on simple aggregate score averages or differences, a few are exploring more complex models that leverage longitudinal data on students to isolate the "value added" by a particular teacher or school. This modeling approach has taken a number of forms and is generally referred to as "value-added modeling" (VAM). Enthusiasm for this approach stems from the belief that it can remove the effects of factors not under the control of the school, such as prior performance and socioeconomic status, and thereby provide a more accurate indicator of school or teacher effectiveness. In this talk, we critically evaluate this claim in the context of models used to isolate teacher effects. We briefly review the principal existing modeling approaches and conclusions from the literature. We then present a general multivariate, longitudinal mixed-model that incorporates the crossed grouping structures inherent to longitudinal student data linked to teachers. We discuss the potential impact of model misspecifications, including missing student covariates and assumptions about the accumulation of teacher effects over time, on key inferences made from the models. We conclude with an assessment of the strengths, limitations, and the most challenging unresolved issues of value-added models as a tool for teacher accountability.
March 5, 2003
The Biostatistician's Role in Health Economics: Adrienne Heerey, Cleveland Clinic Foundation
Rates of healthcare expenditure have risen in excess of inflation in the past decade. The fruits of intensive research in the '80s are becoming realized, resulting in a number of expensive medical interventions becoming available. Simultaneously, focus is being placed on reducing healthcare expenditure. As a result, the requirement for economic evaluation has become vital in the effort to obtain the best value from a given healthcare budget. Due to the scarcity of health economists, many bio-statisticians are being asked to contribute to these studies. This presentation details the fundamental components of cost effectiveness evaluation and economic modeling and highlights statistical queries faced in these studies.

2002 Talks
December 4, 2002
Financial and Estate Planning by the Estate Planning Team, Inc.: Stuart Kleiner, Estate Planning Team, Inc.
Estate Planning is "the action an individual takes to maximize use of their assets to enhance lifestyle and support their desired level of activity, to protect their estate from financial devastation in the event of an unforeseen nursing home stay and to provide for their loved ones through the transfer of property by gifting during lifetime or through inheritance after death, while minimizing all administrative costs and taxes." Mr. Kleiner will discuss basic legal documents every person should have, retirement planning, reducing income and estate taxes, multigenerational family financial planning, college funding, how to help your parents with planning, common planning mistakes, how to select and coordinate professional advisors, etc.
November 6, 2002
Industrial Applications in Statistics: Robert Mason, Southwest Research Institute
Statistics has been used to solve data problems in the United States for over 160 years. The fields of application continue to expand and include common areas such as the physical, engineering, and biological sciences as well as more recent applications such as in space and environmental sciences. This presentation provides an interesting application resulting from recent work conducted at Southwest Research Institute. It concerns a statistical experiment conducted to determine if honeybees can be used to detect land mines. The results have recently appeared in major newspapers, magazines, and network news programs.
October 14, 2002 (Fall Workshop)
A Review: Missing Data: Joseph Schafer, Pennsylvania State University
Statistical procedures for missing data have vastly improved in the last two decades, yet misconception and unsound practice still abound. In this seminar, we will
  • frame the missing-data problem from a statistician's viewpoint
  • introduce fundamental concepts regarding the distribution of missingness and missing at random (MAR)
  • review older, ad hoc procedures including case deletion and single imputation, discussing their merits and weaknesses
  • discuss the theory, implementation and use of maximum likelihood (ML) for incomplete data problems
  • introduce the idea of multiple imputation (MI), discuss its properties, and demonstrate its use on a real data example
  • discuss some of the latest developments in the missing-data literature that have appeared in the last five years, including weighted estimating equations and methods for handling missing values that are not MAR.
September 4, 2002
Optimal Price Rules, Administered Prices and Suboptimal Prevention Evidence from a Medicare Program: Avi Dor, Case Western Reserve University
Pricing methodologies in Medicare vary from one component of the system to another, often leading to conflicting incentives. Failure to recognize linkages between them may result in inefficient allocation of resources and higher overall costs. To motivate the analysis, I derive pricing rules for a welfare-maximizing regulator. I show that while optimal inpatient payments are standard Ramsey prices, optimal outpatient payments must incorporate net loss due to unnecessary hospitalizations, as well as supply elasticities. A myopic regulator will tend to ignore this, leading to underprovision of preventive services. The dialysis program represents a useful case for empirical investigation, since payments for maintenance services more rigidly determined than payments for related hospital care. Given constant prices, empirical analysis focuses on the effect of dialysis intensity on various measures of hospital use. Results indicate that greater dialysis intensity reduces hospital use. Moreover, this is found even at moderate or high levels of intensity, where dialysis is viewed ex ante as being adequate. A simple cost-benefit calculation suggests that for every dollar of additional spending on outpatient intensity, about $1.5 in hospital expenditures can be saved. This suggest that the current pricing structure within aspects of the Medicare program is inefficient, and underscores the problem of regulatory myopia.
June 5, 2002
Assessing Risk and Fairness: The Role of Statistical Science in Policy: David W. Scott, Rice University
Probability theory provides a mathematical framework for modeling risk. Philosophy considers fundamental questions of the nature and meaning of chance. But it falls upon statistical science to collect and analyze data to estimate risk, influence policy, and make decisions. Insurance provides a compelling case study for notions of fairness and subsidy. This talk will examine the notion of a fair game and consider its application in areas including decision making, social security, medical insurance, and exit polling. What are some of the elements of "fair'' policies? A deeper understanding of statistical modeling and evaluation would illuminate subsidies implicit in public policy and would sharpen political debate.
May 1, 2002
Defining 1-Sided P-Values in a Genetic Problem: Robert C. Elston, Case Western Reserve University
Fisher defined the P-value as the probability of an observed result or anything more extreme, i.e. anything that would alert the research worker even more than the observed result to the possibility that the null hypothesis is not true. Depending on the situation, the appropriate P-value should be 1-sided or 2-sided. If segregation at a genetic locus underlies the etiology of a disease, then pairs of siblings who are both affected with the disease will tend to share a larger number of alleles identical by descent at a linked marker locus. Under the null hypothesis that the marker locus is not linked to such a disease locus, the siblings will share 0, 1 or 2 marker alleles identical by descent with probabilities ¼, ½ and ¼, respectively. The definition of a 1-sided P-value in this situation will be discussed.
April 3, 2002
The SAS Solution to Warranty Analysis: Jeff Wright, SAS Institute
SAS is in the process of bringing to market a solution product for Warranty Analysis. This solution has been initially developed for the automotive (and automotive supplier) industry but will soon be applicable to any discrete manufacturer who must both manage warranty claims/costs. The solution will integrate warranty data with key customer, vehicle, production, and geographic information so that the organization can achieve a level of process knowledge that translates into significant value. The solution allows manufacturers to automatically detect emerging quality issues before they make it to the top of the 'issue list'; as needed, use statistical analysis to determine root causes to quickly to focus resources in an accurate and timely manner (using SAS Enterprise Guide, a graphical user interface to the SAS system); as needed, use data mining techniques for further inquiry into manufacturing and customer relationship issues (using SAS Enterprise Miner); forecast warranty costs to protect against financial risk. A demo of the Warranty Solution will be presented along with quick looks at SAS Enterprise Guide and SAS Enterprise Miner.
March 6, 2002
Design of Computer Experiments to Optimize the Mean of a Response: Bill Notz, The Ohio State University
For purposes of this talk, a computer experiment is assumed to consist of computer code that produces deterministic responses for a given set of inputs. We consider the situation in which there are two types of inputs; manufacturing variables and noise variables. Manufacturing variables are those that can be controlled by the product designer. Noise variables cannot be controlled but have values that follow some probability distribution. We seek the values of the manufacturing variables that optimize the mean of a response. The approach is Bayesian; the prior information is that the response is a draw from a stationary Gaussian stochastic process with correlation function belonging to a parametric family with unknown parameters. Following an initial design, a sequential strategy is used to select values of the inputs at which to observe the response. This strategy involves computing, for each unobserved set of values of the inputs, the posterior expected "improvement" over the current best guess at the optimum and the next set of inputs to observe are those that maximize this expected improvement. The strategy is illustrated with examples. Other issues regarding computer experiments will be addressed as they arise.
February 6, 2002
Geometric Morphometric Techniques to Design Skull Prostheses: David Dean, Case Western Reserve University
Recently, we have begun to look at the use of patient specific radiological data as the source for highly accurate computer-aided-manufacture (CAM) of tissue engineered prosthetic implants. The templates for the CAD (Computer-Aided Design) of these prosthetics comes from shape study of human anatomy referred to as geometric morphometrics. Traditional multivariate statistical morphometrics (circa 1880-1980) indirectly gave rise to these methods which are truly a graft from the engineering literature on best fit (rigid body) and warping (forced fit through bending) algorithms. Geometric morphometrics has only truly been a field of inquiry and application for a decade, or so. There are no textbooks, but there are several good conference and review texts as well as journal papers. Most emphasize two primary algorithms, Procrustes and Thin Plate Spine, for best fit and warping, respectively. The primary novelty in these methods is that the transformed landmark coordinates themselves, instead of linear, area, volume, or other traditional measurements, have become the statistical basis of shape measurement. These approaches are useful in describing phenomena of average biological shape or biological symmetry. These data have significant implications for CAD/CAM of human prosthetics.
It is expected to facilitate highly accurate control of implant-host fit and complex internal structures that promote resorption. This work has been done with a 3D Systems (Valencia, CA) SLA 250/40 stereolithographic device. Stereolithographic CAM of PPF components begins with a computer-aided-design (CAD) STL (public stereolithography ASCII data format) file generated in a 3D (i.e., solid) CAD interface. Now that more than 20 patients have received implants produced in this fashion, we are beginning to look at more efficient ways to use stereolithographic production of skull prostheses. Previously other groups and we have used patient CT images as source data to print a skull model of the patient. These large skull models are expensive to produce. Additionally, the manual work that follows to design and manufacture an implant is also expensive. We will present new technology that allows us to print the implant directly thereby saving the cost of printing the patient's skull and expense of several generations of manual work.
January 9, 2002 (Cleveland Chapter Presidential Address)
Analyzing Over-dispersed Count Data from One-Way Designs: Nancy Campbell, John Carroll University
A comparative study of several possible statistical tests for analyzing over-dispersed count data from completely randomized one-way experimental designs will be discussed. Specifically a Monte Carlo study was done in which tests involving the general linear model on the raw data and on transformations of the data were compared to tests based on the generalized linear model utilizing either the Poisson or negative binomial distribution. The problem was motivated by a co-author's frequent encountering of over-dispersed count data from such experiments, often involving small means and small sample sizes. The past simulation study will be explained and conclusions drawn, and a current related comparative study will be discussed as well. In addition, an ongoing undergraduate research project by Computer Science majors at John Carroll done in conjunction with the current study will be discussed.

2001 Talks
December 5, 2001
The Breast Implant Controversy: Statisticians and Epidemiologists Meet the Media and the Legal Profession: W. Michael O'Fallon, Mayo Clinic and Past-President ASA
In 1992, the FDA announced a moratorium on the installation of Breast Implants ending a 30 year period of increasing utilization of this technology. Women with surgically removed or deformed breasts and women who desired breast augmentation had access to such implants in the U.S since about 1962 and were by-and-large satisfied. However, some case reports of a variety of connective tissue diseases occurring in women with implants and a small number of court cases led the FDA to use its newly obtained regulatory authority to request safety data from the various manufacturers of implant materials. When such data were not forthcoming and/or were deemed incomplete or unsatisfactory the moratorium was declared. A firestorm of lawsuits ensued and pleas for scientific information on the topic were issued. The Department of Health Sciences Research at the Mayo Clinic provided the first response with a paper published in the New England Journal of Medicine in 1994. This presentation discusses our contacts with the media and primarily with the legal profession following this publication. Some of our experiences were exhilarating but most were nerve wracking, depressing, disillusioning, and even frightening. However, we have survived.
November 7, 2001
Two-Stage Group Screening in the Presence of Noise Factors: Angela Dean, The Ohio State University
A major advantage of factorial experiments is the information that they provide on interactions. When the number of factors is large and little prior knowledge on the various factorial effects is available, conventional fractional factorial experiments capable of estimating interactions require too many observations to be economically viable. To overcome this problem, interactions are frequently dropped from consideration and assumed to be negligible, often without substantive justification. In industrial experimentation, in particular, the loss of information on interactions is a serious problem, because a key tool for product improvement is the exploitation of interactions between design (control) factors which can be set in the product specification, and noise factors which cannot. Two different strategies for ''group screening'' will be presented for a large number of factors, over two stages of experimentation, with particular emphasis on the detection of interactions between design and noise factors. Three criteria are used to guide the choice of screening technique, and also the size of the groups of factors for study in the first stage experiment. The criteria seek to minimize the expected total number of observations in the experiment, the probability that the experiment size exceeds a pre-specified target, and the proportion of active factorial effects which are not detected. Examples will be used to illustrate the methodology, and some issues and open questions for the practical implementation of the results will be discussed.
October 3, 2001 (Fall Workshop)
Experiments: Planning, Analysis, and Parameter Design Optimization: C. F. Jeff Wu, The University of Michigan
This seminar will be based on the book "Experiments: Planning, Analysis, and Parameter Design Optimization" by Jeff Wu and Mike Hamada (2000). Course notes will be made available to attendees. This book contains many new methods not found in existing textbooks, and covers more than 80 data sets and 200 exercises. The new tools covered include robust parameter design, use of minimum aberration criterion for optimal factor assignment, orthogonal arrays of economic run size, analysis strategies to exploit interactions, experiments for reliability improvement, and analysis of experiments with non-normal responses. Data from real experiments will be used to illustrate concepts. Time will be reserved for questions and discussion.
September 5, 2001
Irregular Factorial Designs: Tena I. Katsaounis, Mansfield Restaurants Inc
Partially Balanced arrays with N runs, m factors, two symbols, strength t greater than two and a given index will be discussed as being Partially Balanced arrays with N runs, m factors, two symbols and of strength equal to two. Formulas for calculating the corresponding index, given any number of factors m, will be presented. This method is useful in showing easily non existence of a Partially Balanced array with N runs, m factors, two symbols, strength t greater than two and a given index. Construction of irregular factorial designs using this method will be discussed. Such designs belong to the class of Partially Balanced or PB1 or Extended PB1 arrays.
June 6, 2001
A Case Study: An Inside Job by an Outsider: Dennis Fox, Cleveland Technical Societies Council President
The presentation will focus on three things. The "Case" study, the results, trends and hopes for these Fourth grade students at a local elementary school. Thoughts on how to make meaningful changes in classes is the second part of the presentation with comments on Time's Schools of the Year Schools That Stretch article in the May 21, 2001 magazine. Finally, a report about a new charter school I hope will use EQL, DDM, QL and SEAQL techniques and other technical societies educational outreach programs as a basis of the curriculum that is scheduled to open this fall in Cleveland. We will be very open to discussion on where and how we go from here at Case Elementary and other programs. Some of the answers, while not difficult or surprising (in hindsight), may surprise you. We are all educators; some of us just received formal training in it.
May 2, 2001 (A Worskshop sponsored by the ASA Council of Chapters)
Permutation Tests: A Guide for Practicioners: Phillip Good, Information Research
This is a one-half day course on permutation methods. It is intended for practicing statisticians and others with interest in applying statistical methods. High school algebra is assumed but no higher-level mathematics is required. Some familiarity with computer simulation would also be helpful. Attendees will be given historical background on resampling methods and a formal introduction to these methods. Emphasis will be placed on the wide variety of applications of the techniques, the computer-intensive nature of implementation along with many examples and "real world" applications. The course is intended for those who use statistical methods in their work. This includes practitioners in medicine, business, engineering and the social sciences. It also will be useful to professors of statistics and those who do statistical research but may not be familiar with resampling methods and want to be updated on the latest advances in methodology and application. Dr. Good is the author of two popular texts on resampling methods. Resampling is a powerful technique which has only recently seen an explosion in applications due to enhancements in computational techniques that make these computer-intensive methods practical.
April 4, 2001
Design of Two-Level Fractional Factorials with Vague Prior Information: Arthur G. Holms
An experimenter's prior information, however vague, is incorporated, according to the vagueness, into the design of expansible sequences of two-level fractional factorials with crossed classification block effects, and all simpler cases. The experimenter's prior subjective probabilities of the importance of model coefficients, block parameters, and likely stopping stages of expansible sequences are used to maximize the expected utility of a design by making optimum selections of model coefficients from the estimates aliased and confounded sets of model coefficients, for all physical to plan variable matchings, for the experiment designer's choices of groups and subgroups of defining contrasts. Differing groups and subgroups can then be compared under the condition of optimal coefficient selection and physical to plan variable matching.
March 7, 2001
Statistics Education and the Making Statistics More Effective in Schools of Business (MSMESB) Conferences: Thomas Love, Case Western Reserve University
The Making Statistics More Effective in Schools of Business (MSMESB) conferences have been held annually since 1986. Since its inception, MSMESB has focused on improving the teaching of statistics and statistical thinking, on cross-disciplinary research, and cross-pollination between academia and industry, and on continuous improvement in business education. The conferences have led to substantial changes in curricular content, modes of delivery, and supplemental material in business (and other) programs all over the world. This talk describes the impact of MSMESB on the teaching of statistics, especially the development of textbooks and software. The talk traces some of MSMESB's history and draws out some practical recommendations. We also highlight some key challenges and opportunities for the future. Earlier versions of this work were jointly prepared with David Hildebrand, of the Wharton School. The MSMESB web site houses details of many of the sessions at http://weatherhead.cwru.edu/msmesb
February 7, 2001
African American Study of Kidney Disease and Hypertension - Design and Update : Jennifer J. Gassman, The Cleveland Clinic Foundation
The African American Study of Kidney Disease and Hypertension is a factorial design study of the effect of antihypertensive regimen and level of blood pressure control on the progression of kidney disease in African Americans with hypertensive nephrosclerosis. In this presentation, the study design will be reviewed and there will be an update on the progress of the study.
January 3, 2001
Uses and Abuses of Proficiency Tests: Robert S. Butler, BFGoodrich
In recent years tremendous emphasis has been placed on the idea of quality in education. To this end many states, including Ohio, have embraced the concept of proficiency testing as a means of quality control. While the pronouncements from Columbus have all of the verbal hallmarks of a quality control effort they have little else. This is unfortunate because, with proper analysis, the proficiency data can provide valuable information about many aspects of the education process. In order to provide an understanding of what can and cannot be expected from proficiency testing, the talk will focus on the analysis of the 1996-1999 Shaker Heights 4th grade proficiency scores.

2000 Talks
December 6, 2000 (Cleveland Chapter Presidential Address)
Cryptography, Primes, and Privacy: Wm. O. Orgel, Solutions; et cetera
There are three words that describe privacy. They are secrecy, secrecy, and secrecy. Article IV of the Bill of Rights guarantees: "The right of the people to be secure in their persons, houses, papers, and effects...". Modern technology is making the privacy of medical records, banking, credit, and our bodies (Genetics), etc. an international issue, which disturbs many people. Cryptography-not thought of by most people very often-is supposed to be at work protecting our privacy every second of every day. Why do we trust cryptology to protect our privacy/security, both personal and national? From whom are we being protected? We will discuss the historical connection among ancient, modern and contemporary mathematics (prime numbers) to the Internet, e.g., Euclid is usually discussed respecting Geometry; and Fermat his Last Theorem, but not this time. There are "keys" to making cryptography work, and we'll discuss them. Is the future of cryptography in Quantum Computing? Finally, we will discuss the Awards of "Privacy International".
November 2, 2000
Statistics for a New Century: Meeting the Needs of a World of Data: Richard Scheaffer, University of Florida, President ASA
The world is awash in data. Most people are aware of the importance and power of data in their professional and personal lives, and many attempt to use data in making everyday decisions. But few are educated in ways that would allow them to comprehend more fully the vast array of uses (and misuses) of data and to use more effectively the quantitative information that confronts them daily. Even fewer are aware of the fact that formal study of statistics can serve to strengthen their own academic preparation for a wide variety of careers. This talk will provide an overview of the current efforts in the United States to infuse statistics into the school (K-12) curriculum and to enhance opportunities for undergraduates to learn more about statistics. Ties to similar efforts in the international community will be mentioned. The goals are to empower students through improved quantitative literacy and to provide strong foundations for careers that depend increasingly on data. Among the strengths of these efforts are the terrific interest they have generated among educators and students at all levels; among the weaknesses are the tendency for programs to become narrow and rote rather than broad and creative. A goal for the next century will be to bring to educational programs in statistics the same kind of creative vitality that marks the practice of statistics among professionals in the field. This will require new emphases in both content and pedagogy. Statistics education has caught the attention of many; it now must prove itself by making effective use of this opportunity to produce new generations of graduates that will not drown in their world of data.
October 4, 2000 (Fall Workshop)
The Grammar of Graphics: Designing a System for Displaying Statistical Graphics on the Web: Leland Wilkinson, SPSS, Inc.
"The Grammar of Graphics" (GOG) is the title of a recent Springer-Verlag book that encapsulates a new theory of statistical graphics. GOG is based on an algebraic model. It contrasts with the prevailing view toward classifying and analyzing charts by their appearance - a view that one might call Pictures of Graphics (POG). In POG, there are pie charts, bar charts, line charts, and so on. Not only are most books and papers on graphics organized by chart type, but so also are most charting programs and production libraries.By contrast, GOG begins with a strong statement: there is no such thing as a pie chart. A pie chart is a stacked bar graphic measured on a proportional scale and transformed into polar coordinates. Significantly, the description of simple charts in POG (such as a pie) are more complex in GOG and seemingly complex charts in POG (such as scatterplot matrices) are simple to describe in GOG. This contrast between surface POG descriptions and deep GOG specifications exposes not only previously unappreciated subtleties in the structures of common charts but also the existence of charts not generally known. GOG is ideally suited for designing a system for interactive Web graphics. Examples will be shown using a Java production library called nViZn (formerly GPL).
September 6, 2000
Epidemiological Causation in the Legal Context: Sana Loue, Case Western Reserve University
Reliance on epidemiological evidence has become increasingly common in various legal contexts, including toxic tort cases, criminal matters, civil lawsuits between individuals for alleged harm, and actions to involuntarily quarantine individuals with specified infectious diseases. However, the operationalization of causation differs between law and epidemiology and the purposes of law and epidemiology are quite different. These divergent methodologies and purposes often result in the misuse and misinterpretation of epidemiological principles and findings in the courtroom. Case examples are used to illustrate these difficulties.
June 7, 2000
Flourishing or Floundering? Relief Pitching in Major League Baseball: Rich Charnigo, Case Western Reserve University
A manager in major league baseball has the often-difficult task of deciding when and how often to switch pitchers during the course of a game. Looking at the simultaneous decline in pitching performance and increase in the use of relief pitching (and perhaps with a few contests from last season in mind), one wonders if relief pitching is being employed past the point of optimal effectiveness. We will consider how relief pitching has changed over the last thirty-nine years and formulate three linear models that will relate the earned run average (a primary measure of pitching performance) to the proportion of complete games (a proxy for the amount of relief pitching). By considering a statistic designed to assess the quality of a pitching change and the proportion of pitching substitutions in 1999 contests that exceed a certain threshold value of this statistic, we will assess the overall effectiveness of relief pitching. With this finding and the results from the aforementioned linear models, we will argue that relief pitching may now be contributing to the very problem which it is designed to circumvent: high scoring by the opposition.
May 3, 2000
Computer Generation of Magic Squares Using Minitab: Josephina de los Reyes, University of Akron
The purpose of this talk is to present a computer method of generating a magic square using MINITAB®, a statistical software. A magic square of order n is an n x n array of the integers 1,2,n2 so that the sum of entries in the n rows, n columns, and main diagonals is a constant. Amid fascinating historical and recreational aspects of magic squares, and theoretical questions asked about it when regarded as a matrix, two ideas in particular caught this author's interest. One idea is the fact that under certain restrictions, a magic square oforder n may be constructed from two mutually orthogonal Latin squares ("MOLS") of order n. Is this procedure reversible for all n - under what conditions can MOLS be obtained from a magic square? Initial results obtained by the author show that given a magic square and a Latin square, another Latin square that is orthogonal is derivable. The second idea comes from an article in electronic engineering about the effectiveness of a dithering algorithm in color printing that is based on a 3x3 magic square over that based on a "direct" dithering algorithm.
April 5, 2000
Computing Multivariate B-splines: A Simulation Based Approach: Nidhan Choudhuri, Case Western Reserve University
Univariate B-splines played an important role, both as a theoretical and practical tool, in dealing with polynomial splines; its multivariate counterpart, despite having all the theoretical properties of the univariate case, lacks application because of the computational difficulties. Unlike the univariate case, where there is an explicit form of the B-spline function, the multivariate B-spline is only implicitly defined and needs numerical approximation. Some computational procedures, based on a recurrence relation formula by Micchelli (1980), are available today. But the computing time there is too long and increases exponentially with the number of knots. In this talk, we shall introduce a new simulation based procedure of computing a multivariate B-spline function, which is less time consuming and easy to implement. Theoretical results will also be presented to support the validity of this procedure.
March 1, 2000
Likelihood Functions, Parametric Statistical Inference and Mathematica: Daniel Cap, DMC Technology
The relationship between parametric probability distributions and likelihoodfunctions plays a fundamental role in both applied and theoretical statistics. The object of this talk is to show how Mathematica, a modern mathematical modeling environment, can be used to explore likelihood functions and make parametric statistical inferences. We'll use simple examples, with binomial, normal and weibull distributed random variables, to illustrate the key ideas. After this introduction, we'll explore practical applications to reliability engineering and robust (Taguchi)engineering design problems, where both the location and scale parameters of the associated likelihood functions are modeled with regression equations.
February 2, 2000
Recent Findings from Two Clinical Surveys: John Holcomb, Youngstown State University
This presentation will discuss two recent projects involving Dr. Ralph Rothenberg, MD of Forum Health Services, Youngstown, OH. The first project involves determining if guidelines mailed to every physician in the country by The American College of Rheumatology impacted baseline and follow-up blood, kidney and liver testing of patients given prescriptions of a non-steroidal anti-inflammatory medication. The second project also involved Dr. Joan Boyd. Here we investigated the effectiveness of health fair osteoporosis screenings. Subjects identified to be at risk for osteoporosis by ankle bone density screening were surveyed six months later to determine if the primary physician was seen and treatment obtained. The talk will present the results of these investigations and thoughts on future research in these areas.
January 5, 2000
The Supreme Court's Decision on the Census: Dan O'Leary, Marconi Medical Systems
On January 25, 1999 the US Supreme Court ruled on a pair of lawsuits relating to the use of statistical methods for apportion of the US House of Representatives. The court ruled that using statistical methods for apportionment of the House of Representatives violates the US Constitution. They also ruled, however, that current law requires the Census Bureau to use statistical methods for other purposes. The presentation reviews the Supreme Court's decision and explains the issues and the Court's reasoning. It looks at the Constitutional requirements for apportionment and the census including a number of constitutional amendments in this area. The presentation touches on the problem of fair apportionment, the legislation relating to the Census, the undercount problem, the initial plan for Census 2000, and the revised plan for Census 2000. Time permitting, we look at some of the arguments for and against sampling.

1999 Talks
December 1, 1999 (Cleveland Chapter Presidential Address)
Acute Care for Elders: Stopping Functional Decline in Hospitalized Elders: Linda Quinn, QED Industries
Patients age 65 and over account for approximately 37% of acute non-federal hospital admissions and 48% of inpatient days in the United States. Functional decline is common in older adults following an acute medical illness and hospitalization. Evidence substantiates that at least one-third of patients age 70 and over lose the ability to independently perform one or more activities of daily living (ADL) after hospitalization for an acute medical illness, and three months post-discharge, 40% of these patients still do not regain preadmission functional status. Patients with ADL decline accompanying an acute illness and hospitalization are more likely to have a prolonged hospital stay, die in hospital, be newly institutionalized at discharge, and be readmitted to the hospital post-discharge. The Acute Care for Elders (ACE) Unit is a multifaceted intervention that integrates geriatric assessment into the optimal medical and nursing care of patients in an interdisciplinary environment. Designed specifically to help patients maintain or achieve independence in self-care activities, the ACE Unit embodies four key elements: a specially designed environment (with, for example, uncluttered hallways, large clocks and calendars, and carpeting); patient-centered care emphasizing independence, including specific protocols for prevention of disability and for rehabilitation; discharge planning with the goal of returning the patient to his or her home; and intensive review of medical care to minimize the adverse effects of procedures and medications. This talk will discuss the series of randomized clinical trials that evaluated the effectiveness of the ACE Unit intervention.
November 3, 1999 (Fall Workshop)
Logistic Regression - A Workshop: Mike Kutner, The Cleveland Clinic Foundation
Data are said to be binary when each outcome falls into one of two categories such as alive or dead, success or failure, true or false. Binary outcome data occur frequently in practice and are commonly analyzed using the logistic regression model. This workshop will emphasize the practical aspects of modeling binary outcome data using logistic regression, including checking the adequacy of the fitted models. Several examples from the health sciences area are presented to illustrate the techniques. Parallels between multiple linear regression modeling and multiple logistic regression modeling will be presented throughout the workshop. Therefore, attendees of this workshop are expected to be familiar with multiple linear regression modeling techniques.
October 6, 1999
Optimal Small Response Surface Designs: Tena Katsaounis, MS
Three level factorial designs will be discussed that are suitable for small second order designs. The best choice of potential design points will be discussed under the criterion of minimizing the generalized variance of the parameters in the model. Two methods will be illustrated that yield designs with different resolution for the main effects and two factor interactions. Both methods involve the use of Partially Balanced and PB1 arrays. The Extended Partially Balanced Array will be defined as a generalization of PB.
September 8, 1999
Handicapping Systems for Sailboat Racing: Paul Mathews
Wherever there are sailboats in the water there are sailboat races. There are almost as many race scoring systems as there are different kinds of sailboats. When boats that are racing are all the same design (one design racing) or are very similar in design (box rules) then scoring is easy and accurate. However, most racing fleets contain a mixture of boats. In a single race it's not unusual to find boats that run from 20 to 50 feet long, that weigh from 2500 to 30,000 pounds, that race with 3 to 15 people, and that cost from $3000 to $300,000. Under these diverse conditions performance handicapping systems have evolved so that everyone has a chance (or is supposed to have a chance) to win. Paul will use video clips to demonstrate the scale of the handicapping problem. He will discuss the committee driven handicapping systems that are based on elapsed time (Time-On-Time) and course length (Time-On-Distance) and the ratings systems that are based on aero- and hydrodynamic computer models (velocity prediction programs). Then he will present a new rating system that is based on the results of computer physical modeling but is refined with actual race data. Paul will also talk about some of the technological developments that will affect rating systems in the next five years.
June 2, 1999
Optimal Experimental Design in Poisson Impaired Reproduction Studies: Jennifer Huffman, The Lubrizol Corporation
Impaired reproduction experiments are a class of studies involving Poisson responses that are functions of concentrations of toxicants, chemotherapy drugs, or other substances. Factorial-like D, Ds, and interaction optimal designs are demonstrated for models involving interaction through k factors. Augmentations of these designs that result in desirable lack of fit properties are discussed. The augmentation contains "center runs" that are analogous to standard center runs in factorial designs for linear models. In addition, fractional factorial designs are detailed along with their alias structure. Robustness properties are addressed as well.
May 5, 1999 (Joint Dinner Meeting at the University of Akron)
Proportional Hazards Models With Informative Censoring: P.V. Rao, University of Florida
A modification of the proportional hazards model is proposed for informatively or randomly censored times to two types of events -- a primary event and a followup event. The proposed model treats informative censoring as a type of risk in a competing risks setup and incorporates suitable parameters to describe the conditional probability of failure given the subject is not informatively censored. Inferences about the parameters in the model can be based on standard partial likelihood theory. The model is used for analyzing some data on waiting time to bone marrow transplantation (primary event) and time to relapse or death after transplantation (followup event) for leukemia patients.
April 7, 1999
Careers in Statistics: the Real World A Panel Discussion: Susan Cowling, The Lubrizol Corporation; David Nelson, The Cleveland Clinic; Patricia Cirillo, Cypress Research Group
The panel presentation will be from 4:00pm to 5:00pm. The panel consists of three local statisticians who are employed in different application areas. They will discucss: What do I do? What types of analyses do I run? What type of degrees are required/useful? What skills are important? Salary and Job market?
March 3, 1999
A Unified Regression Approach for Fixed Effects ANOVA Models with Missing Cells: Mike Kutner, Cleveland Clinic Foundation
Many standard statistical software programs claim to "handle" fixed effects ANOVA models when some of the cells do not contain data (missing cell problem). Simple examples using SAS, SPSS, and BMDP will be given documenting how these programs "handle" the missing cells. Surprisingly, these computer packages generally produce strange results. The sources of these difficulties will be delineated. A unified full-rank regression cell means model approach will be presented as a definitive solution to this problem. Experimental design models, such as the balanced incomplete block design, will be shown to be special cases.
February 3, 1999
Using Economic Designs for Multivariate Control Charts: Thomas Love, Case Western Reserve University
The economic design of control charts means making rational decisions about several parameters, including control limits, sample (subgroup) sizes, sampling intervals, and degree of data smoothing in light of a model for the costs and time associated with monitoring and controlling a process. Technological advances have sparked interest in multivariate process control methods - which allow for the simultaneous study of several quality characteristics. The Multivariate Exponentially Weighted Moving Average (MEWMA) control chart is especially attractive to practitioners looking for methods that effectively detect small shifts in a process mean vector. We discuss the use of economic designs for multivariate control charts, focusing on the MEWMA. This presentation is the result of joint work with Kevin Linderman, at the University of Minnesota.
January 6, 1999
"Actuarial" vs. "Actual" Analyses: Accounting Simultaneously for Multiple Competing Risks of Events: Eugene Blackstone, Cleveland Clinic Foundation
Typical life tables (generically called "actuarial" tables, be they related to survival of patients or reliability of machines) address the distribution of times to a single event. In some settings, however, other events may occur prior to the event of interest that effectively remove a patient or item from the possibility of experiencing the event of interest. For example, one may wish to study the need for additional procedures following a heart bypass operation. However, death may occur before the need for an additional procedure. This usually does not affect the estimates of the probability of reintervention, but it does affect the "actual" number of reinterventions that will be observed in patients across time. Or a heart valve manufacturer may know that there is a high probability that a valve will deteriorate over a period of many years, but if these are used in elderly people, the number of actual valve replacements for deterioration may be low because of natural attrition from other causes. In the late 18th century Daniel Bernoulli presented the mathematics for multiple competing events, using it to predict the impact on survival of a population were smallpox to be eradicated. This led to multiple decrement life tables in demography and the general theory of competing risks in statistics. This talk will be introductory in nature using buckets with different sized leaks as a motif, providing motivation for the technique in several medical settings, examining the appropriateness of the methodology to different questions being asked, and briefly introducing both nonparametric and parametric estimation methods.

1998 Talks
December 2, 1998 (Cleveland Chapter Presidential Address)
Five Randomized Clinical Trials of Abciximab During Percutaneous Coronary Intervention: Shelly Sapp, Cleveland Clinic Foundation
Platelet thrombus formation, resulting from platelet activation, adhesion and aggregation, predisposes patients to the development of ischemic complications after percutaneous coronary interventions (PCIs). Clinical evidence suggested that sustained platelet inhibition after PCIs may reduce the occurrence of post-hospital events such as myocardial infarction and coronary revascularization. Recently, a new class of platelet antagonists directed against the platelet membrane glycoprotein IIb/IIIa receptor has undergone extensive clinical testing. One of these agents, Abciximab (ReoPro, Centocor), blocks this receptor, thus preventing platelet adhesion and aggregation. Between 1991 and 1998, five randomized, placebo-controlled clinical trials of Abciximab during percutaneous coronary intervention involving a total of 9038 patients have been completed. This talk will describe the motivation behind the design of these trials as well as some of the major results.
October 29, 1998 (Dinner meeting)
Geographical Trends in Cancer Mortality: Using Spatial Smoothers and Methods for Adjustment : Karen Kafadar, Ph.D., University of Colorado-Denver
Mapping health-related data can lead to important insights and generate hypotheses about causes and potential effects. Such data are commonly adjusted for the variables age and gender, so that inferences are not influenced by these obvious factors. In a similar fashion, data for certain diseases ought to be adjusted for known risk factors. One method of adjustment is suggested here, and insights from the adjusted data are enhanced by smoothing the data in the two dimensions (longitude and latitude). The process of adjustment and smoothing is illustrated on three sets of cancer mortality data: lung cancer (using urbanicity as the adjustor), prostate cancer in nonwhites (using percent African-American as the adjustor), and melanoma among whites (using latitude as the adjustor). In each case, the maps of the adjusted rates indicate patterns that are worthy of investigation and may contribute to the generation of hypotheses and further understanding of the etiology of the diseases.
October 7, 1998
Testing For Equivalence of Diagnostic Tests: Nancy Obuchowski, Ph.D., Cleveland Clinic Foundation
Studies comparing the diagnostic accuracy of clinical tests are common, particularly in the field of radiology. Accuracy is defined in terms of the test's sensitivity, specificity, or indices associated with the Receiver Operating Characteristic (ROC) curve. In comparing two tests, we are often interested in determining if a new test has similar accuracy as an existing test, i.e. `Are the two tests equivalent?'. We propose two criteria for defining diagnostic equivalence and methods for testing equivalence. The criteria are referred to as `Population Equivalence' and `Individual Equivalence'; they are modifications of criteria used for assessing equivalence between generic and standard drugs. The proposed methods are illustrated for a study comparing the diagnostic accuracy of digitized mammographic images to original film. The digitized images are easy to store and retrieve, but the digitization process may result in a loss of accuracy. We test whether the accuracy of the digitized images is equivalent to film and whether the management of individual patients will be impacted if digitized images replace films.
September 9, 1998
Is Rank Transformation Method a bad idea?: Guang-Hwa "Andy" Chang, Ph.D., Youngstown State University
The rank transformation (RT) refers to the replacement of data by their ranks, with a subsequent analysis using the usual normal theory procedure, but calculated on the ranks rather than on the original data. This idea was originally suggested by Lemmer and Stoker (1967) and advocated by Conover and Iman. The availability of statistical packages for parametric tests makes the rank transformation method appealing. SAS had also added this option in their package. However, Blair, Sawilowsky and Higgins (1987) showed that, for 4x3 factorial designs, a severe inflation in Type I error of the RT statistics for testing interaction is observed as either the cell size becomes large or the row and column main effects are large. It was a huge disappointment. Is the rank transformation method a bad idea? Some research results after the simulation study by Blair et al. will be presented in this talk.
June 3, 1998
Genetic Mapping of Complex Human Diseases: Jane M. Olson, Ph.D., CWRU
In recent years, genetic study of complex human diseases has increased dramatically. Most human diseases are now believed to have some genetic component, and considerable effort is being made to find and study the genes involved. As a result, statistical methods used to find disease genes are receiving a great deal of attention, and statistical mapping methodology is evolving rapidly. In this talk, I will provide an overview of genetic mapping methods, focusing primarily on genetic linkage analysis. I will first explain concepts in genetic inheritance that statisticians exploit in genetic mapping, and introduce relevant terminology. I will then explore the two main types of genetic linkage analysis: model-based and model-free. In model-based linkage analysis, one estimates a genetic model of inheritance for the disease using pedigree likelihood methods, then fixes the model in subsequent estimation of the linkage parameter that describes the relationship between the inheritance of disease and marker loci. In model-free linkage analysis, a smaller set of parameters that are functions of the genetic model and the linkage parameter are estimated, so that the genetic model of the disease need not be known prior to analysis. Instead, sharing of marker alleles between related individuals is exploited. I will discuss the relative usefulness of these approaches in practical genetic mapping problems, and provide some discussion of future directions.
May 6, 1998
Good Apple? Sampling Biases: Jiayang Sun, Case Western Reserve University.
In practice, data are more often from "hell" thaom "heavn fren". They may come with missing values, censored or truncated observations, outliers and/or from a biased sample. If missing values are missing at random (and are not too many), deletion or an imputation procedure may be used to clean up the data before analyses. If whether a data point is missing depends on the true value, the data have come from a biased sample. In the presence of sampling biases, standard procedures often fail badly. In this talk we present some fun examples with conclusions that were drawn ignoring biases, and illustrate why some cases with censoring or truncated observations may be considered from a biased sampling. We then offer some simple solutions and go into the author's current research in this area.
March 24, 1998 (Joint Dinner meeting with Case Western Reserve University)
Follow-up Designs that Make the Most of an Initial Screening Experiment: Robert Mee, Department of Statistics at University of Tennessee in Knoxville.
Industrial experimentation is often sequential, with initial screening experiments followed by other stages of experimentation. Common alternatives for subsequent experimentation include the path of steepest ascent, augmenting an initial fractional factorial via foldover, and adding axial and center points to complete a central composite design. This talk will focus on two other alternatives: augmentation via semifolding and non-central composite designs. In each case we find opportunity to use estimates from the initial experiment to direct location of the follow-up design.
March 4, 1998
Navigating the Net: Susan E. Branchick, Ricerca
The Internet contains a wealth of information, but locating it can be a frustrating experience. This talk will present tips and techniques to make your experience on the Internet a smoother ride. Topics will include how to subscribe to a discussion group, work within frames on the WWW, search for information and handle documents. It will conclude with an overview of some statistical related sources.
February 4, 1998
Statistical Literacy and Statistical Competence in the 21st Century : David S. Moore, 1998 President, American Statistical Association
Educated people face a new environment at century's end: work is becoming intellectualized, formal higher education more common, technology almost universal, and information (as well as mis- and dis-information) a flood. In this setting, what is statistical literacy, what every educated person should know? What is statistical competence, roughly the content of a first course for those who must deal with data in their work? If the details are automated, are the concepts and strategies that guide us enough to maintain "statistics" as a distinct discipline?
January 7, 1998
Statistical Opportunities and Job Search Fundamentals: Greg Jarecki, Trilogy
This presentation will focus on three areas: career trends in the statistical field, elements of successful career planning, and fundamentals of job hunting. Included with career trends are career opportunities, skill requirements, and salary ranges for various statistics professionals. Included with job hunting fundamentals are resume preparation, and the differing perspectives of job candidates and employers.

1997 Talks
December 3, 1997 (Cleveland Chapter Presidential Address)
Quantitative Investing: Bill Katcher
Is the stock market just another process? Yogi Berra says, "You can see a lot just by looking." I believe that a lot can be learned with basic statistical methods like regression analysis, data plotting, and time series analysis. Spreadsheets are the tool that make it possible for the individual investor to monitor the process in near real time. There are two parts to the investment decision. First, what's the overall market direction and second, what specific investments are the best at this time. I hope to be able to give you some new ideas in both areas.
November 5, 1997
Designing for Better Data, Timely Analysis, and Effective Reporting: Mark Martin, Chiron Diagnostics Corporation
As statisticians, we too frequently encounter the frustration of spending more time than we like cleaning and preparing data. Ideally, we would like to spend the majority of our time designing for, analyzing, and reporting from good data. Doing so requires a good balance of foresight, collaborative planning, information sharing, tools, continuous learning, and communication skills. Experiences from a fast-paced project will be shared, including successes and pitfalls. While specific practices vary in different environments, certain basic principles apply across workplaces. An extensive bibliography of good references (for both statisticians and non-statisticians) will be provided. As a specific tool, the use of "data flow diagrams" will be demonstrated.
October 1, 1997
Rank Analysis of Means: Dan Sommers, General Electric
The typical non-parametric procedure for analyzing a one-factor design is the Kruskal-Wallis test. An alternative procedure, using a rank approach to the one-factor analysis of means, will be presented.
September 10, 1997
Testing for Treatment Effects which are Increasing Functions of the Rate of Disease Progression: Tom Greene, Cleveland Clinic Foundation
Clinical trials in chronic disease populations are often designed to test whether a treatment intervention slows disease progression as assessed by the slope of repeated measurements of a marker of disease severity. For continuous markers, mixed effects models are generally used to test for additive treatment effects on the mean slope, implicitly assuming that the treatment effect for each patient is independent of that patient's progression rate. However, a close examination of the hypothesized mechanisms of the treatment often suggests that the clinical hypothesis can be better formulated by a treatment effect whose magnitude increases as a function of the progression rate. We introduce a mixed effects model in which the treatment effect is proportional to the rate of progression for patients with negative slope, and equal to zero for patients whose slope is non-negative. An estimated least squares approach is proposed for estimation and hypothesis testing in this setting. We show that use of standard additive analyses can lead to markedly lower power than the proposed procedure when the treatment effect increases with the effect size. We also examine time-to-event analysis based on the time to a specified clinically meaningful reduction in the outcome marker as a practical alternative to mixed models when increasing treatment effects are considered likely. These approaches are illustrated with examples from clinical trials in renal disease.
June 4, 1997
Projection Methods for Generating Mixed-Level Fractional Factorial and Supersaturated Designs: Alonzo Church, Jr.
The definitions of resolution and projectivity have been used to develop an algorithm to find mixed-level fractional factorial designs. Some of the designs differ from standard designs and have superior projection properties. In addition their least squares properties are often superior. In his presentation, Mr. Church will describe the algorithm and give details on some useful alternative designs.
May 7, 1997
Forensic Economics: John F. Burke, Ph.D.
In his talk entitled "Forensic Economics", Dr. Burke will concentrate on the ways statistics and probability can be statistical evidence in jury trials, he will use actual cases involving racial discrimination and the asbestos industry as examples.
April 2, 1997
Alternative Control Charts: Tom Ryan, Case Western Reserve University
Traditional methods for determining control chart limits have appealed to the Central Limit Theorem when an Xbar chart is used and to the normal approximations to the binomial and Poisson distributions when attribute charts are used. Unfortunately, these approaches can produce very poor results. Recent research on improved techniques for measurement and attribute data will be presented, including methods developed by the speaker. The control chart panel discussion that is appearing in the April issue of the Journal of Quality Technology will also be briefly discussed.
March 5, 1997
Cleveland Technical Societies Council: Fred Lisy, CTSC
The Cleveland Technical Societies Council (CTSC) brings together the members of the technical, and engineering societies in Northeastern Ohio to: Provide a forum for discussion and a vehicle for action on matters of mutual concern and interest; Serve as a focal point for contact between industry and society at large with the technical and scientific community, Promote intersociety communications, professional interchange of ideas and coordination of society activities; Promote interest in and encourage careers in the scientific and technical professions through career guidance and other educational programs. The CTSC has approximately 50 member societies and associations such as American Association of Cost Engineers (NE Ohio Section), American Institute of Aeronautics and Astronautics, American Institute of Chemical Engineers, the American Society for Information Science, the American Society for Quality Control (Cleveland Section), the Cleveland Computer Society, the Cleveland Engineering Society, Society of Fire Protection Engineers and the Cleveland Chapter of the American Statistical Association. The CTSC program's include: Technology infrastructure such as educational programs complement, education program database, career day, science fair mentoring and local student programs; Technology transfers such as consortiums and R&D partnerships issues, technology policy forum and regional engineering technical symposiums; Technology development such as forums for technology awareness, public awareness of technology and industry-government R&D partnerships.
February 5, 1997
Update on the Data Coordinating Center and the African American Study of Kidney Disease and Hypertension: Jennifer Gassman, Cleveland Clinic Foundation
The African American Study of Kidney Disease and Hypertension (AASK) is a randomized clinical trial sponsored by the National Institute of Diabetes, Digestive and Kidney (NIDDK) Diseases of the NIH. There are 21 Clinical Centers across the United States, including one in Cleveland at Case Western Reserve University. The Data Coordinating Center is in the Deparment of Biostatistics at the Cleveland Clinic. Statisticians working on the study include Mike Kutner (Principal Investigator), Jennifer Gassman, Tom Greene, and Shin Ru Wang. In this 3x2 factorial design study, participants receive 1 of 3 blinded antihypertensive agents (ACE inhibitors, beta blockers, or calcium channel blockers) and either usual blood pressure control (MAP of 102 to 107 mm Hg) or low blood pressure control (MAP <92 mm Hg), where MAP = 1/3 systolic + 2/3 diastolic blood pressure. The primary outcome variable is rate of change in glomerular filtration rate (GFR), a measure of kidney function. Data are entered directly from the clinical centers into a central Oracle database, using the Internet. The study has been going on for about two years, and there are about four more years to go. In this non-technical presentation, the speaker will give an update on what the Data Coordinating Center is doing and how the AASK Study is progressing.
January 8, 1997
Response Surface Methods for Costly Noisy Data: Art Holmes
The presentation will include a quick review of the concepts of response surface methods. The review will be followed by the rationale of a specialization to high efficiency central composite experiments for costly noisy data. For 2 through 8 independent variables, design tables will present the efficient numbers of center points that can be distributed among the blocks. The numbers of center points vary from 2 to the number of hypercube blocks plus 4. The tables list the star point radii for orthogonal blocking.

1996 Talks
December 4, 1996 (Cleveland Chapter Presidential Address)
Presenting Results - Using Tabular Displays Effectively: John Schollenberger, Ricerca
Graphical displays are important, but, we also need to give careful consideration to our use of tables. The design and layout of the tabular displays in a report can also help or hinder a client's understanding of the results. My recent experiences suggest that, more often than not, tables are used only to provide a listing, either of raw data or of summary statistics and results, with little thought as to their digestability. That is, we don't consider how that table, or a different table, might be better used to assist the reader in understanding the results or believing the conclusions that are drawn. I will step through a suggested redesign of several tables.
November 6, 1996
Inference About the Change-Points in a Sequence of Random Vectors: Arjun K. Gupta, Bowling Green State University
In this talk, the change-point problem is reviewed. Then, testing and estimation of multiple covariance change points for a sequence of m-dimensional (m>1) Gaussian random vectors by using Schwarz information criterion (SIC) is studied. The unbiased SIC is also obtained. The asymptotic null distribution of the test statistic is also derived. The result is applied to the weekly prices of two stocks (m=2), Exxon and General Dynamics, from 1990 to 1991, and changes are successfully detected.
October 2, 1996
The Meaning of Analysis of Means: Edward G. Schilling, Rochester Institute of Technology
The Analysis of Means is becoming increasingly well known. It is a method for the graphical analysis of the averages of proportions resulting from a designed experiment.The procedure has been applied to a variety of experiments including crossed and nested fixed effects models, balanced incomplete blocks, split-plot, etc. It is especially useful as a communication tool in an industrial environment since it is based on Shewhart control chart concepts. The background and use of this approach will be discussed in terms of industrial examples.
September 4, 1996
Constructing a Deterministic Model of an Epidemic Process When All Parameters Must be Estimated : John Neill, the City of Cleveland Department of Public Health
It is often impossible to obtain the parameters required to construct a classical deterministic model of an epidemic process since the number of people at risk and contact rate are usually unknown. New diseases, such as AIDS, may have no known incubation period. A technique is presented for using observed incidence to estimate all the parameters necessary to make deterministic models of diseases like AIDS or hepatitis A. The method also produces estimates of incubation distributions for AIDS and hepatitis A comparable to those in the medical literature.
June 5, 1996 (Joint Meeting with the Cleveland SAS Users Group)
Statistical Quality Control Using the SAS System : Dennis King, STATKing Consulting
This presentation will be divided into two parts. The first part discusses the seven tools of quality with emphasis on the SAS software and programming useful for implementing these tools. The second part of the presentation focuses on the control chart and the statistical methodology surrounding the use of this tool.
May 1, 1996
Pros and Cons of Cutpoints: Jane Goldsmith, Health Sciences Biostatistics Center at the University of Louisville
Authors have noted the disadvantages of dichotomizing or taking cutpoints in the data. Some textbooks and educators still advise students to dichotomize in order to use convenient Chi-square statistical analysis. Statistical packages make this transformation easy. This talk summarizes the costs of dichotomization, considering efficiencies of Chi-square versus t-tests, Mann-Whitney U tests, Regression and Correlation Analysis, and Spearman Correlation Analysis. Practical examples will be given for cases when dichotomization is desirable and undesirable from a power standpoint. Discussion of logistic regression and implications for the use of CART are included.
April 3, 1996
Statistician Meets the Customer: Mark Martin, Ciba Corning
On an automated immunochemistry system which runs more than 25 different blood tests, customer satisfaction can vary with the performance of the assays that a particular hospital or laboratory runs on the instrument. During assay development, one challenge is to predict the customer complaint rate if the assay were to be released at a given point in time, and to provide indicators of when an assay meets release-for-sale goals that will please the customer. As a part of this effort, interviews were conducted at several customer sites.
March 6, 1996
On The Importance of Assessing Measurement Reliability When Fitting Regression Models : Leon J. Gleser, University of Pittsburgh
In many regression contexts, some predictor variables are measured with error, or replaced by proxy measurements for reasons ranging from convenience to cost. In such cases, adjustments to classical least squares slope estimates may be needed to correct for bias. These adjustments require knowledge about the parameters of the model that is often not supplied by the data. It is argued here that the required knowledge concerns the reliability matrix of the vector of measured predictor variables. Ways to design auxiliary experiments to obtain reliability information, and to combine such information with information available from published studies, are discussed.
February 7, 1996
A Paradigm Shift in Statistical Consulting: Mukul Mehta, Quality Sciences, Inc.
Eighties show TQM emerge as a major movement in American industry. The TQM wave is now followed by reengineering and corporate right sizing. Staff groups are out of fashion or under intense pressure to produce more with less. This is an excellent opportunity for an industrial statistician to prosper or perish. Time to market is becoming a dominant force in the market place and will forever continue to be. For professional statisticians to survive and prosper we need to rethink how TQM reengineering and time to market affect our profession and what we need to do to respond to challenges. We need a paradigm shift. Information technology is becoming the key component of reengineering and has the potential of forever changing the traditional approach to statistical consulting. QSI has developed a software-based approach to enable large groups of scientists and engineers to take advantage of the power of statistical thinking without the pain. Through slides and software, Mukul will illustrate issues, challenges and opportunities that lie ahead.
January 10, 1996
Making Statistics Understandable for Students and Industry: Paul Mathews, Lakeland Community College
Paul's students come from industry, most with 5 to 20 years experience as machine operators, technicians, or supervisors. They have a growing interest in quality engineering but, unfortunately, very limited math skills. They quickly come to the opinion that people write textbooks to make money (based on their experience in the bookstore) and impress their friends. Paul will describe some of the notational conventions he has found well received by the students, and show some of their favorite graphical presentations.

1995 Talks
December 5, 1995 (Cleveland Chapter Presidential Address)
Portrait of a Cleveland Clinic Foundation Statistician: Lisa Rybicki, The Cleveland Clinic Foundation
George Williams arrived at the Cleveland Clinic Foundation in August 1980 and the Department of Biostatistics was born. George began to expand the department in 1981. Now, 15 years later, the department employs 80 people, almost half of which are statisticians. The mission of the department is to excel in the conduct of medical and methodological research, and to promote the proper use of statistics and epidemiological methods. The department fulfills this mission through a wide variety of collaborative and educational activities. Lisa will describe the organizational structure and work environment, illustrate the mechanisms used to fulfill their mission, and examine the role of the collaborative biostatistician in this dynamic environment.
November 1, 1995
Implementation of Total Quality Management in an R&D Environment: Dennis J. Keller, RealWorld Quality Systems
This talk gives an overview of the unique challenges and solutions associated with implementing Total Quality Management (TQM) in a technical or R&D environment. It begins with defining TQM, its goals and objectives, and how these differ from their well understood counterparts in manufacturing and general business arenas. Next, a generalized global framework of hierarchical "systems" is presented, which when taken as a whole, represents the "System" of TQM in R&D. Each system in the hierarchy is then explored in only moderate detail to demonstrate its critical function in the global system. The system hierarchy are: 1) Project/Problem Solving, 2) the Researcher's Personal TQM System, 3) the Branch/Group/Division TQM System, and 4) the overall TQM System. Lastly, a short overview will be given on strategies for actual implementation and managing for change.
October 11, 1995
Product Optimization in the Chemical Additives Industry: Carlos L. Cerda de Groote, The Lubrizol Corporation
Very often in industrial practice there is an interest in understanding how to control and optimize a response. Two optimization examples are given. The first example deals with formulating practice. One wants to know how a blend's composition affects viscosity. The goal is to obtain cost effective blends that satisfy viscometric requirements. Predictive viscometric equations are fitted to data and then used together with constraints and a cost function to obtain minimum cost formulations. The second example deals with a predictive model for the molecular weight distributions (MWD) of blends of polymers. Engine performance depends on the MWD of the polymer(s) used with the additive chemistry. Control of this parameter also leads to opportunities for optimization within the formulation framework. In both examples the optimization was done with Microsoft Excel.
September 6, 1995
News and Numbers: How Reporters Can Tell Facts From Lies, How You Can Help: Victor Cohn, Research Fellow of ASA
Journalists and the public are often confused by constantly conflicting "they say's" about health, science, the environment and many other subjects. How can you help reporters report facts, or the best facts you can muster, correctly? Tell them about uncertainty. All you can give them is the best estimate at the moment. Tell them about the use of probability, the power of large numbers, the pitfalls of variability and the dangers of bias. Some basic rules for dealing with the media and informing the citizenry will be discussed. Remember that candor builds credibility.
June 7, 1995
Statistics in the Insurance Industry: James Jiang, Progressive Insurance
Insurance rates are mainly determined by the past loss experience. Statistical methods have been employed to analyze the historical data to set up adequate rates. even though there are some differences in reasoning between a casualty actuary and a statistician, they often arrive at similar conclusions. Following an introduction to automobile insurance rate making, James will demonstrate how statistical science can be applied to reach some important results in credibility theory and risk classification. The alternative reasoning by the casualty actuaries will also be presented.
May 3, 1995
Robust Methods for the Detection of Linkage and the Identification of Genes for Complex Disorders: Deborah V. Dawson, Case Western Reserve University
Genetic linkage refers to the tendency of two different genes to be inherited together if they are located sufficiently close together on the same chromosomal strand of DNA. Modern molecular biology has put at our disposal a vast array of genetic markers, located at intervals throughout the human genome. The establishment of genetic linkage between a known marker system and a putative gene for a disorder is generally regarded as the ultimate statistical evidence for a genetic component in the disease etiology. Statistical methods for the detection of linkage include maximum likelihood based approaches and the so-called robust or model-free approaches. The latter are usually based on pairs of relatives, and in contrast to the likelihood based methods, do not require specification of the mode of inheritance for the disease trait.
April 12, 1995 (Joint dinner meeting with the ASQC)
Quality is Personal: Harry V. Roberts, University of Chicago
Personal quality is the application of quality ideas to your job (and even to your life). Personal quality does not directly tell you how to do your job better. Rather, it offers a philosophy and methodology for learning how to do it better. Personal quality helps people to perceive and remove waste from their own jobs. As a result, they can achieve both continuous improvement and breakthroughs. Some potential gains to the individual care increased job satisfaction, education in quality management concepts and tools, and assessment of the usefulness of quality management concepts based on direct experience rather than the claims of others. Personal quality differs from traditional time management although it may help to make better use of time management tools. Personal quality is not just another self-improvement program like speed reading, public speaking, or memory development, although these skills can lead to improvement of personal quality.
March 1, 1995
Statistical Graphics: Innovate! Show the Data: Ralph O'Brien, The Cleveland Clinic Foundation
Today's computing tools allow us to create statistical graphics that are customized to the particular hypotheses, research designs, measures, and conclusions at hand. This demands higher levels of "statistical graphicacy" from both authors and readers. We should be open to innovation and tolerant of some complexity. As Tufte proclaimed, "Above all else, show the data". In clinical trials, for example, we must strive to show how patients respond to treatments, so that we can see both general tendencies and individual variation. These points will be discussed by studying examples.
February 1, 1995
Statistics and the Law: Lynn D. Sivinski, Case Western Reserve University
In spite of the apparent disparity between the two fields, there are several areas where statisticians can assist lawyers in litigation. Lynn will discuss a few of these subject areas and how the statistical tools are used in making arguments. She will also discuss how legal standards of proof and scientific ones differ.
January 11, 1995
Quantitative Literacy: A Data Driven Curriculum: Jeff Witmer, Oberlin College
One of the many spin-offs of ASA's Quantitative Literacy program is the Data Driven Curriculum Project. The goal of this project, which is supported by an NSF grant, is to develop materials for mathematics courses taken in grades 7-12. These materials are intended to be used within existing courses in algebra, geometry, etc. The Cleveland Chapter gas been very active in supporting the QL program for several years and Jerry Moreno has planned a 1995 summer workshop based on the Data Driven Curriculum materials. During this talk we will look at the project and some of the activities within it.

1994 Talks
December 7, 1994 (Cleveland Chapter Presidential Address)
The Ethical Statistician: Linda Quinn, Case Western Reserve University
A working definition for unethical professional behavior as a statistician is any action intended to mislead, misinform, or mask information without making clear the relevant limitations of the outcome and its presentation. From incorrect analyses and violated assumptions to fraudulent data and misrepresented results, ethics and statistics interact.
November 2, 1994
Interim Analysis for Clinical Trials: A Generalized Spending Function Approach: Hussein R. Al-Khalidi, The Proctor & Gamble Company
In most clinical trials, patient recruitment occurs over a period of years and data or information from the trial accumulates steadily throughout it's duration. The trial should involve the smallest number of patients required to reach a firm decision. Group sequential methodology in which groups of patients are analyzed periodically is an approach that is economical, ethical and allows for steady accumulation of data. Boundaries for maintaining a fixed significance level will be discussed. Since interim analyses will be conducted at times other than at those planned, a proposed generalized spending function will be used to generate discrete boundaries. The optimality property of the proposed function will be discussed.
October 5, 1994
Beyond the Shewhart Paradigm: J. Stuart Hunter, Princeton University
In today's 'bits and pieces' assembly industries modern instrumentation allows the precise measurement of almost every item produced. The consequence is time series data not unlike that obtained in the continuous process industries. The basic concepts underlying the popular Shewhart charts are thus often inappropriate when applied in an assembly industry environment. This lecture proposes an alternative, yet simple model that can extend the usefulness of the Shewhart methodology and, if desired, lead to immediate active control of the industrial process.
September 7, 1994
Robust Data Analysis Made Simple: Tom G. Filloon, The Proctor & Gamble Company
In many practical situations an observed data set may contain extreme values (heavy tailed/contamination) that may cause one to be leery of normality based conclusions. Hence a robust analysis may be in order. However, a more efficient analysis might be obtained by something other than a log or rank transformation approach. Tom will briefly outline various robust methods and then discuss M estimation in some detail. Implementing a pseudo value approach, he will show how M estimation can be carried out via simple least square (such as in SAS) and the utility of this method will be demonstrated using several real examples.
June 1, 1994
Some Interesting Regression Modeling Examples: Mike Kutner, The Cleveland Clinic Foundation
As a consulting statistician one sometimes comes across data sets that are not routine textbook examples. Three such examples will be presented and solutions proposed. Each of these examples are from "real life" data sets. In addition, all three examples are relatively small data sets with only two to four potential predictor variable that need to be modeled.
May 4, 1994
Reporting and Reviewing Research Results: Thomas Lang and Michelle Secic, The Cleveland Clinic Foundation
No one can deny the existence of a long standing problem in the medical literature; errors and omissions in reporting statistical results. The speakers are developing a comprehensive set of guidelines for presenting statistical information in biomedical publication entitled Reporting Statistical Information in Biomedical publications: A Guide For Authors, Editors and Reviewers. The purpose of the guide is to aid in recognizing, understanding a properly reporting statistical analyses. The guide is being developed from a thorough review of the literature. It will not serve as a statistics text but will function as a reference designed to help authors, journal editors, medical students and peer reviewers understand and correctly present statistical information.
April 13, 1994
Practical Issues and Difficulties in Coordinating a Stroke Treatment Trial - A Statistician's Perspective:Robert Woolson, University of Iowa
In Prof. Woolson's talk a particular trial is discussed with emphasis on issues such as distributed data entry, reliability assessment, training an monitoring.
March 2, 1994
Getting Started with PROC MIXED: Kristen Latour, The SAS Institute
This presentation introduces PROC MIXED using examples that include split-plot designs, repeated measures data and one way ANOVA with unequal variance. The GLM and MIXED procedures are compared in terms of syntax, output and interpretation. An overview of SAS statistical modeling procedures in included.
February 2, 1994
Multivariate Data Analysis with New Visualization Methods:Vladimir Grishin, Case Western Reserve University
Experienced statisticians know that the effectiveness of statistical methods does depend on the choice of solution structure or data model, feature selection, results interpretation, etc. Today they solve these problems more logically by applying knowledge, experience and sometimes looking at visual displays which are often difficult to interpret. Data analysis by pictorial representation is based on years of research and modeling of human vision abilities to effectively detect and describe complicated nonlinear dependencies directly in large dimensional data space.
January 5, 1994
Applications of Statistics and Operation Research Methods:Mike Sweeney, American Greetings Corporation
Applications of statistics and operations research methods at American Greetings Corporation will be discussed. A background on the company will be presented and a description of how Management Sciences services the needs of the various internal departments will be provided. Concentration is on solving business problems through an blend of pure statistics/O.R. and general analytical methods.

1993 Talks
December 1, 1993 (Cleveland Chapter Presidential Address)
Comparing the Robustness of Alternative Methods of Exploring Response Surfaces: Gary Skerl, Glidden Paints
Traditional methods of response surface analysis frequently involve fitting a quadratic model to a central composite or Box-Behnken design. This approach purports to give a reasonable approximation to the "true" response surface is more complex. New approaches are emerging which claim to perform better. Two of these approaches include neural networks and locally weighted regression. A protocol is proposed for comparing the robustness of these alternative. An example is given using a simulated test case.
November 3, 1993
Optimum Experimental Designs for Comparative Bioavailability