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).
Big Data & Rare Events: Standard Fixes Can Backfire
Insights from the Field
separation
rare events
dyadic data
cauchy prior
Methodology
PSR&M
1 archives
Dataverse
Separation and Rare Events was authored by Liam Beiser-McGrath. It was published by Cambridge in PSR&M in 2022.

Separation problems in binary models can mislead even large datasets if both dependent and independent variables are rare, such as interstate relations data. Researchers often use Firth's penalized likelihood approach to overcome this issue, but our simulations show it may lead to incorrect inferences. An alternative solution involves Bayesian methods with weakly-informative priors centered at zero—specifically the Cauchy prior—to avoid these pitfalls.

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