Title
Recovery of Density Function Information from Regression Equations when Dependent Variables are Polychotomous
Speaker
Thomas W. Blaine, Extension Service, The Ohio State University
Abstract
Researchers are often in a position of attempting to measure attributes of a density function without being able to observe the variable in question directly (like willingness to pay for a program designed to achieve environmental improvement), but only some other variable associated with it. Typically, the observed variable comes in a limited form however (thumbs up or down on a referendum for example). This has given rise to a wide variety of procedures not only to estimate moments associated with the density function in question, but also to estimate how the target variable is influenced by exogenously determined variables (like income). Results obtained from logistic, ordered probit and ordinary least squares (OLS) regressions are presented and discussed. Avenues of future approaches statisticians should consider in addressing this topic are explored.
Meet the speaker in Room 212 Cockins Hall at 4:30 p.m. Refreshments will be served.
