Informal Seminars Presented In Autumn 2002

 

October 2, 2002

 

Title: 

 

Speaker:

Ruth Etzioni, Ted Gooley, Yutaka Yasui

Abstract: 

In an effort to get to better know some of the faculty and learn about research that is going on in the department, we've invited 3 faculty members from Fred Hutchinson Cancer Research Center to speak to us.  They will talk about some of their current research interests and statistical problems they're involved in.  As many students are in the process of picking thesis advisors and many faculty are looking for students, we're hoping this will be a helpful seminar.

 

 


 

October 9, 2002

 

Title: 

 

Speakers:

Carolyn Rutter, Charles Kooperberg

Abstract: 

This is the second of two special seminars aimed at helping us meet some of the faculty who we don't see often around the department. Three faculty members will discuss their current research.  Nancy Temkin is a joint faculty member with Neurological Surgery, Carolyn Rutter is with the Center for Health Studies at Group Health, and Charles Kooperberg is with Fred Hutchinson Cancer Research Center.   As many students are in the process of picking thesis advisors and many faculty are looking for students, we're hoping this will be a helpful seminar.

 

 


 

October 16, 2002

 

Title: 

 

Speaker:

David Lockhart, Graduate Student, Biostatistics

Abstract: 

David Lockhart will be giving the informal seminar talk tomorrow.  He will be discussing statistical issues that arose in the analysis of an observational study examining neuropsychological (NP) function and HIV-RNA levels in blood and cerebrospinal fluid (CSF) among 25 patients at 3 time points before to two months after initiating highly active antiretroviral therapy (HAART).  Results for the full 25 subjects yielded inconsistent associations at different time points; however, it appears this may be to a change in the study protocol which introduced confounding between time and the use of antiretroviral mono- or dual- therapy (ART) at baseline..  A subset analysis of those who were not on ART at baseline revealed an association betweeen decreased CSF viral load and increased NP scores, but the same associations were weaker and inconsistent for decreased blood viral load and NP scores. Patients in this subset taking regimens including drugs with good CNS penetration showed more improvement in NP scores, but did not have larger drops in CSF viral load, than those whose regimens did not contain any drugs.  These results suggest that the monitoring of CSF viral load and the use of CNS penetrant drugs may improve management of HIV-related dementia, though further work is necessary to demonstrate the clinical significance of these effects.

 

 


 

October 23, 2002

 

Title: 

 

Speaker:

Anton Westveld Graduate Student, Statistics

Abstract: 

Decision making under uncertainty has long been an area of interest to a

wide variety of academic disciplines: statistics, computer science,

philosophy, political science, economics, and biology, to name a few. The

main mathematical method for examining multi-agent decision theory has been

game theory. However, the game theoretic solutions of some simple games have

been called into question, with a classic example being the Sub-Game Perfect

Nash Equilibrium (SPNE) of the centipede game. In experimental settings,

individuals rarely choose the SPNE solution (McKelvey and Palfrey (1992)).

An explanation of this has been the notion of learning. As individuals play

the game, repeatedly, they will move toward the SPNE (Fudenberg and Tirole

(1991)). An extensive amount of theoretical work has been written on the

subject, as well as, a limited amount of empirical work (El-Gamal, McKelvey,

Palfrey (1993)). Recently, a general class of statistical models has been

developed allowing for a statistical interpretation of game theoretic models

(McKelvey and Palfrey (1996)). Within this framework, it is possible to

examine the statistical evidence for learning in games.

 

 


 

October 30, 2002

 

Title: 

Of Mice and Men (and Cats)

Speaker:

Bryan Shepherd Graduate Student, Biostatistics

Abstract: 

Bryan Shepherd will be giving the informal seminar today.  His talk is

titled "Of Mice and Men (and Cats)".  He will be speaking about work he

has been doing to estimate the replication and differentiation rates of

human hematopoietic stem cells.  Hematapoietic stem cells are the cells

that give rise to blood cells.  These parameters are fairly well known in

mice and cats, but the methods for identifying them in mice  and cats

cannot be ethically performed on humans. Using relevant data from mice,

cats, and humans, stochastic models have been created aiming have been at

identifying plausible values for the replication and differentiation rates

in humans.  In this presentation, Bryan will discuss these models and

some of the statistical issues that have arisen.  This work is part of his

RA with Peter Guttorp in the Statistics Department and Jan Abkowitz in the

Hematology Department.

 

 


  

November 6, 2002

 

Title: 

Differential Item Functioning

Speaker:

Lance Jolley Graduate Student, Biostatistics

Abstract: 

Tests with individually scored questions, or items, are commonplace

in medical, educational, and sociological settings.  To rely on the

results of these tests, it is important to be certain that the tests are

free from bias to relevant subgroups.  Differential Item Functioning (DIF)

occurs when subjects from different groups have differing probabilities of

success on an item, after controlling for overall ability.  Ways to detect

DIF will be discussed with an emphasis on ordinal regression.  Multiple

comparison issues will also be addressed.

     My group has developed an interactive STATA program to detect DIF.

The Cognitive Assessment Screening Instrument (CASI) used for dementia

screening is used as an example.

 

 


  

November 13, 2002

 

Title: 

Some research topics in health services research and diagnostic medicine

Speaker:

Andrew Zhou

Abstract: 

In this talk, I will present a description of my current research in

health services research and diagnostic medicine. I will describe some research

problems in analyses of health care costs, causal inferences in observational

studies and randomized trials with non-compliance, non-parametric estimation of a receiver operating characteristic (ROC) curve, and correction for

verification bias in ROC curve studies.  Finally, if I have time, I will also

discuss some opportunities for biostatisticians in health services research and

clinical studies that are being conducted at the VA Puget Sound Health Care

System.

 

 


  

November 20, 2002

 

Title: 

Logistic Regression with Covariate Measurement Error:  Estimation and a New Measurement Model

Speaker:

Elizabeth Sugar, Graduate Student, Statistics

Abstract: 

One of the main focuses of Epidemiological research for the last several

decades has been the relationship between nutrient consumption and disease

risk.  For example, several studies have been done to try to determine

whether or not there is a direct association between dietary fat intake

and breast cancer but there have often been conflicting results.  One

possible cause for such discrepancies is the presence of measurement error

in the covariates.  When attempting to deal with this problem we must

address two issues: parameter estimation and modeling for the error

structure.  I will give a brief background of the parameter estimation

techniques which are commonly used along with a brief comparison of

these methods.  I will then then present our new measurement

error model and discuss the extension of the estimating techniques to

this new error structure.  If there is time I will briefly discuss the

problem of sandard error estimation for the parameters.

 

 

November 27, 2002

 

Title: 

Logistic Regression with Covariate Measurement Error:  Estimation and a New Measurement Model

Speaker:

Sierra Li, Graduate Student, Biostatistics

Abstract: 

One of the main focuses of Epidemiological research for the last several

decades has been the relationship between nutrient consumption and disease

risk.  For example, several studies have been done to try to determine

whether or not there is a direct association between dietary fat intake

and breast cancer but there have often been conflicting results.  One

possible cause for such discrepancies is the presence of measurement error

in the covariates.  When attempting to deal with this problem we must

address two issues: parameter estimation and modeling for the error

structure.  I will give a brief background of the parameter estimation

techniques which are commonly used along with a brief comparison of

these methods.  I will then then present our new measurement

error model and discuss the extension of the estimating techniques to

this new error structure.  If there is time I will briefly discuss the

problem of sandard error estimation for the parameters.

 








Last Modification: 4 December  2002