A Regression Model Building Strategy for Confirmative Analysis


Details
Abstract: In data analyses, regression models are used for one of three purposes: descriptive, predictive, and confirmative. The overarching goal of scientific research is to confirm a hypothesis based on a dataset and a reasonable model based on scientific rationale. In this talk, we focus on confirmative regression modeling with multiple variables. There are many popular variable selection methods (e.g., forward, backward, stepwise selection), but there is no alternative for thinking about scientific question, so we want to avoid automatic procedures. Confirmatory data analysis must be hypothesis-driven rather than data-driven. In this talk, we discuss a model building strategy for confirmative regression model (i.e., hypothesis testing). In particular, we discuss why we adjust variables and what variables to adjust when we investigate an association between an explanatory variable and a response variable.
Speaker: Steven Kim
Parking info:
The easiest parking option is to park in lot 71 on Inter-Garrison Rd. at the intersection of Inter-Garrison and 6th Ave. Unfortunately, you will need to pay for parking, using the yellow box that you see as you enter the lot.

A Regression Model Building Strategy for Confirmative Analysis