Abstract

The logit-logistic distribution is a distribution on the unit interval, and could be considered an
alternative to the beta distribution. It apparently first appeared in a 1982 paper by Tadikamalla
and Johnson. We give a definition of the distribution and show some of its basic properties. We
then use both the beta distribution and the logit-logistic distribution to model a clinical data
set in which the dependent variable is a proportion. The performance of the two models is
compared. A final example will be given, showing an unexpected use for the logit-logistic
distribution.