5. An application to regression modelling
5.1. A suitable reparametrization
Gómez-Déniz et al. (2014) gave a reparametrization of the
Log-Lindley distribution with (σ , λ)-parametrization in order to
provide an alternative to the beta regression model, appropriate for situations where the dependent variable is bounded (see Ferrari and Cribari-Neto, 2004). The reparametrization was obtained by
equating its mean to a new parameter θ , leading to