Statistics Colloquia

Spring 2019

Future colloquia are currently being planned. Please revisit this page for further updates.


September 28, 2018

Bo Li, PhD (University of Illinois at Urbana-Champaign)

2 - 3 p.m. (DU 212): Spatially Varying Autoregressive Models for Prediction of New HIV Diagnoses

Abstract: In demand of predicting new HIV diagnosis rates based on publicly available HIV data that is abundant in space but has few points in time, we propose a class of spatially varying autoregressive (SVAR) models compounded with conditional autoregressive (CAR) spatial correlation structures. We then propose to use the copula approach and a flexible CAR formulation to model the dependence between adjacent counties. These models allow for spatial and temporal correlation as well as space-time interactions and are naturally suitable for predicting HIV cases and other spatio-temporal disease data that feature a similar data structure. We apply the proposed models to HIV data over Florida, California and New England states and compare them to a range of linear mixed models that have been recently popular for modeling spatio-temporal disease data. The results show that for such data our proposed models outperform the others in terms of prediction.


September 21, 2018

Peter Müller, PhD (University of Texas Austin)

2 - 3 p.m
. (DU 276): The Future of Bayesian Clinical Trial Design

Abstract: The notion of one treatment serving a large homogeneous patient population is becoming increasingly hard to sustain.  Many recent studies are designed to understand and address heterogeneity of patient populations, exploiting features of adaptive treatment allocations, population enrichment and sequential stopping. In an increasing number of studies the discovery of relevant subpopulations for such adaptive treatment is part of the trial design. We review some novel clinical trial designs that implement such schemes, using examples with increasing levels of adaptation.  First, we start the discussion with adaptation based on a patient's first cycle response in a two-cycle treatment.  Next we continue with dynamic treatment regimens that include adaptation on the outcome from the initial front-line therapy. The discussion includes an adjustment for lack of randomization in the assignment of later stage salvage therapies.  Third, we review a basket trial design for a study of targeted therapies for cancer. In this study adaptation includes the selection of disease, treatment and a patient subpopulation.  Common to these examples is the notion of quantifying the value of alternative treatment allocations and outcomes. In all examples we do this using a utility function that formalizes, for example, the tradeoff of toxicity and efficacy outcomes.  A last example shows another application of such utility-based designs. This time without the context of adaptation.  A common theme in the examples is the use of model-based methods for statistical design and inference, also known as Bayesian methods. 

4 - 5 p.m. (DU 276): Bayesian Categorical Matrix Factorization via Double Feature Allocation

Abstract: We propose a categorical matrix factorization method to infer latent diseases from electronic health records data. A latent disease is defined as an unknown cause that induces a set of common symptoms for a group of patients. The proposed approach is based on a novel double feature allocation model which simultaneously allocates features to the rows and the columns of a categorical matrix. Using a Bayesian approach, available prior information on known diseases greatly improves identifiability of latent diseases. This includes known diagnoses for patients and known association of diseases with symptoms. We validate the proposed approach by simulation studies including mis-specified models and comparison with sparse latent factor models. In an application to Chinese electronic health records (EHR) data, we find results that agree with related clinical and medical knowledge.

View past colloquia >>