253 – Contributed Oral Poster Presentations: Biometrics
The Hosmer-Lemeshow Goodness-of-Fit Test: Does the Grouping Really Matter?
Zoran Bursac
University of Arkansas for Medical Sciences
Dave Hosmer
University of Massachusetts
Hillary Rivera
University of Arkansas for Medical Sciences
Keith Williams
University of Arkansas for Medical Sciences
The Hosmer and Lemeshow goodness-of-fit statistic (HLGOF) is a well established method of assessing model fit for logistic regression. This statistic is based on the grouping of predicted probabilities. It is available in many software packages and routines that offer logistic regression modeling. Some suggest that the performance of HLGOF depends on the decile grouping that the test statistic is based on. We propose a modification to HLGOF decile risk approach via random grouping method. In this project we describe our approach and discuss its performance through simulations.
