Title: Statistical Modeling of Human Fecundity with a view towards individualized risk prediction for pregnancy

Speaker: Rajeshwari Sundaram, Senior Investigator, Biostatistics and Bioinformatics Branch, Division of Intramural Population Health Research, Eunice Kennedy Shriver National Institute of Child Health and Human Development

Human fecundity, defined as the biologic capacity of men and women for reproduction irrespective of pregnancy intentions, is of considerable public health interest, given growing evidence supporting its worldwide decline.  Some researchers argue that environmental reproductive and/or developmental toxicants are responsible for the decline while others posit that behavior is the culprit, given that couples are intentionally postponing childbearing.  As clinical and public health groups work toward the implementation of preconception clinical guidance for couples at risk for pregnancy and more globally reproductive health, it is imperative to understand both male and female determinants of fecundity and fertility.  To accomplish this goal, statistical models that incorporate biological features of both members of the couple are needed.  This is challenging since complex multivariate longitudinal modeling techniques need to be developed to adequately reflect the biology. In this talk, I will focus on statistical modeling for predicting time-to-pregnancy and for understanding the effects of important behavioral factor, ie intercourse on time-to-pregnancy.  I will discuss recently developed statistical models ranging from discrete survival time as well as models for day specific probabilities of pregnancy that have been developed to address the analytical challenges mentioned above while relaxing some implausible biological assumptions that existed in prior approaches.  Lastly, I will discuss methods that allow us to assess some easily available longitudinal processes assessing behavior and menstrual cycle characteristics and show how they can be used to predict time-to-pregnancy dynamically. All these approaches will be illustrated on data arising through various prospective pregnancy study designs.

Title: Control-based Imputation for Missing Data Handling in Longitudinal Clinical Trials

Speaker: G. Frank Liu, Ph.D., Merck Research Laboratories, North Wales, PA

[Abstract] Control-based imputation, an approach which imputes missing data in the test drug group using a model built from the control group, has gained more attention in recent research for handling missing data in clinical trials. This control-based imputation (CBI) approach typically provides a conservative point estimate for treatment difference, and addresses an estimand which has some causal-effect interpretation. A standard multiple imputation approach with Rubin’s rule is commonly used to implement this method. However, the combined variance from Rubin’s rule may over-estimate the variance, therefore reduce power for treatment comparison in statistical inference. In this talk, we discuss some alternative methods on getting more appropriate variance for CBI analysis in different types of outcomes from longitudinal clinical trials including continuous, binary, and recurrent time to event. Applications to several real clinical trial datasets are presented for illustration.

11/8/2017 Wednesday

Title: Inference in Increasing Dimension

Speaker: Fang Han, Department of Statistics, University of WashingtonDepartment of Statistics

Abstract: Since 1960s, statisticians have started realizing the importance of characterizing the role of parameter dimension in statistical inference. For this, pioneers like Peter Huber started introducing a new feature into asymptotics: the number of parameters, p, is now allowed to increase with the sample size n. Following this track of thinking, in 1980-90's, statisticians like Stephen Portnoy and Enno Mammen developed a set of seminal results, sharply characterizing the limiting behaviors of regression estimators and exploring the growth rate boundary of p with regard to n. More recent results, aiming at tackling more general statistical problems, include He and Shao (2000) and Spokoiny (2012). In this talk, I will give a systematic review of these results, as well as disclose more new features. In particular, simple criterion will be given for guaranteeing asymptotic normality and bootstrap consistency for general problems under the boundary condition p^2/n -> 0. The proof rests on Talagrand generic chaining for empirical processes and Gine-Zinn multiplier inequality. Smoothness will be shown, as always, to play the key role in analysis.