Thomas Lumley • 1/27/2017

Case-control sampling and pseudo-Rsquareds

This technical article addresses the challenge of computing pseudo-R-squared statistics (Nagelkerke, Cox-Snell, Tjur) for logistic regression models fitted to complex survey data, such as case-control designs. It explains how standard calculations are biased under informative sampling and proposes design-based versions using survey weights to estimate population-level model fit, with links to a preprint and code.

0 comments

#statistics #Logistic Regression #Case Control Study