statrefs home‎ > ‎Main‎ > ‎Methods‎ > ‎Regression‎ > ‎Logistic Regression‎ > ‎

Firth Penalized Likelihood

Useful in cases of complete or quasi-complete separation. 

For the theory behind the method, see the attached article '[Biometrika] Bias reduction of ML estimates (Firth).pdf' and other related information attached below.

Available in R using package 'logistf'.  Also see package 'zelig'.

Available in SAS 9.2.  See for SAS macro.