STA 701 Talk

Speakers:  Bo Liu and Jennifer Kampe
Moderator: Vivek Singh

Speaker: Bo Liu

Title: Power and sample size calculation for propensity score analysis of observational studies

Abstract: Power and sample size calculations in causal inference with observational data are increasingly desired. Propensity score is one of the most widely used methods in causal inference, but there lack related power calculation tools.  This paper develops theoretically justified and interpretable analytical strategies for power calculation in propensity score analysis. We introduce an easily computable measure of covariate overlap between treatment groups. To address the central challenge of lacking individual covariate information prior to the intended study, we propose to use a linear combination of the multivariate covariates to model the treatment assignment and outcome generating mechanisms that satisfy pre-specified data features. We develop an associated R package \texttt{PSpower} and illustrate the method and software with real world examples.

Speaker: Jennifer Kampe

Title: Nested exemplar latent space models for dimension reduction in dynamic networks

Dynamic latent space models are widely used for characterizing flexible changes in networks and relational data over time. These models assign each node a vector of latent attributes characterizing connectivity with other nodes, with these latent attributes changing dynamically over time.  Node attributes can be organized as a three-way tensor with modes corresponding to nodes, latent space dimension and time. Unfortunately, as the number of nodes and time points increases, the number of elements of this tensor becomes enormous, leading to substantial computational and statistical challenges, particularly when data are sparse. We propose a new approach for massively reducing dimensionality by expressing the latent node attribute tensor as low rank. This leads to an interesting new nested exemplar latent space model, which characterizes the node attribute tensor as dependent on low-dimensional exemplar traits for each node, weights for each latent space dimension, and exemplar curves characterizing time variation. We study properties of this framework, including expressivity, and develop efficient Bayesian inference algorithms. The approach leads to substantial advantages in simulations and applications to ecological networks.