Student Academics Club Seminars

The BSO is starting a new student academic club! The club will meet for a one hour seminar each month where we will hear talks that are given from other students. The talks or presentations will be given strictly by students and will provide good examples of how students in our department achieve academic success and build a platform to enhance academic communication between students and faculty.

The academic club will have the following types of presentations: student publications, conference talks, software workshops, journal presentation, literature searching workshops and interview talks.

Any ideas for presentations or software workshops are certainly welcomed. If you are interested in giving talk(s) or you are looking forward to hearing talks from a specific area, please let us know.

BAYESIAN HIERACRHICAL MODELS for MULTI-CENTER CLINICAL TRIALS: POWER AND SUBGROUP ANALYSES

04/25/2008

Emine Ozgur BAYMAN

Abstract:

The first part of the presentation is about the power and sample size calculations for multi-center clinical trials. An important issue in a multi-center clinical trial is to decide the total sample size and also the impact of any imbalance in the sample size for each center, especially when there is variability between centers.  A concept of Bayesian power is defined as the probability of reaching a specific posterior conclusion for a fixed set of parameter values.  An algorithm is described for determining the sample size to give appropriate Bayesian power. Corresponding frequentist power is calculated, by fitting a GLMM model.  The overall power and type I error rates with both Bayesian and frequentist analysis are compared. 

The second part of the presentation is about detecting the qualitative interaction with a Bayesian approach. Analysis by center in a multi-center trial is a subgroup analysis.  Differences between treatment effects between centers in a multi-center trial represent interaction.  Peto (1982) defines quantitative and qualitative interaction.  Quantitative interaction occurs when the simple effects for treatment in each subgroup category are different and the signs of the treatment effects are identical across all subgroups.  Qualitative interaction occurs when the simple effect of treatment in at least one subgroup is in a different direction than in other subgroups.  A Bayesian test of qualitative interaction is developed by calculating the posterior probability of qualitative interaction and Bayes Factor.  The frequentist power and type I error of the Bayesian test is examined and compared to other approaches such as the method of Gail and Simon (1985) and Piantadosi and Gail (1993).

OPTIMAL ADAPTIVE GROUP SEQUENTIAL DESIGN FOR PHASE II CLINICAL TRIALS: A BAYESIAN DECISION THEORETIC APPROACH

04/04/2008

 Yiyi Chen

Abstract:

Bayesian decision theoretic approaches have been widely studied in the literature as tools for designing and conducting phase II clinical trials. However, full Bayesian approaches that consider multiple endpoints are lacking. Since the monitoring of toxicity is a major goal of phase II trials, we propose an adaptive group sequential design using a Bayesian decision theoretic approach, which characterizes efficacy and toxicity as correlated bivariate binary endpoints. We allow trade-off between the two endpoints to a certain extent. Interim evaluations are conducted group sequentially, but the number of interim looks and the size of each group are chosen adaptively based on current observations.

We utilize a loss function consisting of two components: the cost associated with accruing, treating, and monitoring patients, and the loss associated with making incorrect decisions. The operating characteristics of the design are evaluated over a range of parameter values assigned to the loss function.

Our method is illustrated in the context of a single-arm phase II trial of bevacizumab, gemcitabine, and oxaliplatin in patients with metastatic pancreatic adenocarcinoma.

An adaptive, two-phase, dose exploration design for the estimation of a Human Colonizing Dose 50 (HCD50) and Human Colonizing Dose 90 (HCD90).

04/04/2008

 Yu-Hui H. Chang

Abstract: 

An adaptive design was developed to inoculate healthy volunteers with nontypeable Haemophilus influenzae.  The ultimate goal of estimating the doses at which 50% and 90% of subjects become colonized (the HCD50 and HCD90).  This study was designed to guide in the design of subsequent studies and to explore the dose response relationship.  A fifteen-subject study was designed in two stages, with the first six subjects allocated sequentially.  The design was chosen based on heuristic arguments.  The performance of this design is evaluated by simulation, under both Bayesian and frequentist criteria.  The alternative designs are Bayesian and use myopic methods.  Results from the original study are used to design a subsequent study to better estimate the HCD50 and HCD90 using Bayesian methods.

Job Interview Skills for Biostatisticians

02/22/2008

Dr. Jeffrey D. Dawson

Abstract:

As a faculty member at Iowa since 1991, I have observed the concerns and uncertainty (and sometimes the fears) that arise when our students start looking for jobs.  When I teach introductory classes to our majors, I try to emphasize the fact that biostatisticians should be involved in all steps of the research process, because our jobs involve much more than statistics.  Similarly, the topics covered in a job interview often extend beyond the scope of technical skills that have been covered in biostatistics/statistics courses.  I will demonstrate this with one or two mock job interviews, followed by a discussion

Is Your CV ready?

11/30/2007

Qian Cicci Shi

Abstract:

This week we are lucky to have Cicci share her experience in job hunting with us.  Cicci will start her job at Mayo clinic in Spring, 2008.  In her talk, she will tell us about the job market for newly graduated biostatisticians.  She will talk more about something she thinks she did well, something that she definitely massed up, something she think she should have done but she didn’t, some small tips for the trips, and some tough questions she still remember and feel scared, …..  

This will be a useful talk for all of us. Knowing what skills and abilities our future employers are looking for will better prepare us for the job market when we graduate.  It is always good to be well prepared and early prepared.

Extending Regression Models using Penalized Approaches

11/16/2007

Breheny, Patrick J

Abstract:

Regression models have been among the most important statistical tools for decades.  As the collection, storage, and transfer of information becomes cheaper, however, problems involving large data sets with dozens/hundreds/thousands of potentially important predictors are increasingly common, which poses a number of problems.  A very useful approach to estimation in these situations is to incorporate penalties into the objective function (e.g. squared error, log likelihood) that a regression seeks to minimize/maximize.  In this talk, I will introduce the idea of penalized regression and discuss three important instances: ridge regression, the Lasso, and penalized regression splines.  An in-depth treatment of all the computational, distributional, and theoretical issues related to penalized regression could fill an entire course, so this talk will focus on the conceptual and applied aspects.  At the conclusion, an example analyzing factors influencing attendance in a public health intervention using these methods will be presented.