FIND DATA: By Author | Journal | Sites   ANALYZE DATA: Help with R | SPSS | Stata | Excel   WHAT'S NEW? US Politics | Int'l Relations | Law & Courts
   FIND DATA: By Author | Journal | Sites   WHAT'S NEW? US Politics | IR | Law & Courts
If this link is broken, please report as broken. You can also submit updates (will be reviewed).
Insights from the Field

Does Your Interaction Effect Really Need a Product Term? New Evidence on Logistic Regression


product term
logistic regression
interaction testing
conditional effects
Methodology
PSR&M
1 archives
1 text files
Dataverse
Compression and Conditional Effects: A Product Term Is Essential When Using Logistic Regression to Test for Interaction was authored by Carlisle Rainey. It was published by Cambridge in PSR&M in 2016.

This paper argues that political scientists incorrectly theorize interaction effects due to variable compression alone, leading to biased results if product terms aren't included in logistic regression models. While simulation studies show models with product terms fit non-interactive relationships surprisingly well, removing the bias toward finding spurious interactions is possible by including this term.

data
Find on Google Scholar
Find on JSTOR
Find on CUP
Political Science Research & Methods
Podcast host Ryan