Please note that for the next few weeks we are having 2 seminars per week (Tues & Thur). Below is the information for next weeks 2 seminars. Please also note the location of the seminars.
Statistical Inference for Diagnostic Classification Models, Gongjun Xu (Columbia University)
Speaker Gongjun Xu, Columbia University
Date January 22, 2013
Time 4:00 pm – 4:50 pm
Location 165 Everitt
Sponsor Department of Statistics
Event type Seminar
Abstract:
Diagnostic classification models (DCM) are an important recent development in psychological/educational testing. Instead of an overall test score, a diagnostic test provides each subject with a profile detailing the concepts and skills (often called “attributes”) that he/she has mastered. Central to many DCMs is the so-called Q-matrix, an incidence matrix specifying the item-attribute relationship. It is common practice for the Q-matrix to be specified by experts when items are written, rather than through data-driven calibration. Such a non-empirical approach may lead to misspecification of the Q-matrix and substantial lack of model fit, resulting in erroneous interpretation of testing results. This talk is concerned with data-driven construction (estimation) of the Q-matrix and related statistical issues of DCMs. I will first give an introduction to DCMs and an overview of recent developments, followed by a discussion of key issues and challenges. I will then present some fundamental results on the learnability of the Q-matrix, including sufficient and necessary conditions for it to be identifiable from data. I will also present a data-driven construction of the Q-matrix and estimation of other model parameters, and show that they are consistent under certain identifiability conditions.
******************************************************************************************************************************
Exploring Spatial Heterogeneity in Species Prevalence, Avishek Chakraborty (Texas A&M University)
Speaker Avishek Chakraborty, Texas A&M University
Date January 24, 2013
Time 4:00 pm – 4:50 pm
Location 269 Everitt
Sponsor Department of Statistics
Event type Seminar
Abstract:
In present day ecological studies, complex models for analyzing species abundance patterns are gaining importance because of their multiple utilities — to understand the response of the species to climate variation, to predict its prevalence in remote areas, to measure the impact of human activities and, most crucially, to design effective strategies for conservation. The Cape Floristic Region (CFR) in South
Africa, a well-known biodiversity hotspot, provides a rich class of such species data for analysis. However, any usual regression model would be inadequate due to multiple sources of bias and imprecision in the data — irregular sampling effort, unobserved explanatory features, removal of forest cover, missing a presence and over/under reporting of abundance concentration. Additionally, depending on how the data have been recorded, the response can be a complete collection of abundance counts (presence-absence data) or just a set of locations where the species has been observed (presence-only data). In this talk, I am going to present flexible spatial models based on hierarchical Bayesian approach that can account for these challenges leading to meaningful inference. Given the fine scale resolution of the observed variation over a typically huge study region, implementation of such models can be resource and time-expensive. I shall discuss ways of efficient parallelization or approximation within the estimation scheme, in context of these models that make it feasible to carry out the computation in reasonable time. Further analysis of such datasets presents significant research opportunities in developing models for flexible correlated processes as well as in addressing questions relevant to ecology and environmental science.