Graduate Student Seminar

Speaker 1: Sai Popouri

Title

An Expectation-Maximization (EM)-like method for daily precipitation data using covariate daily series

Abstract

We will start with a brief discussion of the Expectation-Maximization (EM) method of estimation. We will discuss an application of the EM-like algorithm in modeling the daily precipitation series at a location in the Missouri River Basin (MRB) using the daily precipitation simulated by a Global Climate Model (GCM) as a covariate. The covariate daily series is used to develop a transfer function in the lagged regression context. Concepts from time series analysis will be presented as needed.

Speaker 2: Zhou Feng

Title

Estimate Causal Effect under Noncompliance

Abstract

Randomized clinical trials that compare treatments are generally straightforward to analyze. However, when some participants do not comply with their assigned treatments, the analysis is complicated. One popular estimator used for such trials is intention to treat (ITT) estimator, which actually provides a measure of the effect of treatment assignment by classifying participants according to their assigned treatments, regardless of whether they complied with the treatment.  One estimand of the effect of treatment itself is complier-average causal effect (CACE), which is the average treatment effect in the subpopulation of principal compliers.Based on certain settings, three simple estimators, per-protocol (PP) estimator, instrumental variable (IV) estimator and as-treated (AT) estimator, will be introduced to estimate the effect of treatment.