The logit exponentiated power exponential regression with applications

Abstract

We introduce a new distribution, called the logit exponentiated power exponential, defined on the unit interval. Explicit expansions are derived for its moments. Also, we propose a regression based on this distribution with two systematic components, which can provide better fits than the beta and simplex regressions. Its parameters are estimated by maximum likelihood. Some simulations investigate the accuracy of the estimates. The usefulness of the new models is proved by means of three real data sets.