A bootstrap estimator for the Student-t regression model

The Student-t regression model suffers from monotone likelihood. This means that the likelihood achieves its maximum value at infinite values of one or more of the parameters, in this case the unknown degrees of freedom. This leads to problems when one uses iterative algorithms to locate the solutions to the non-linear equations generated by the likelihood. Fonseca et al. (2008) deal with this problem by using the Jeffreys priors. We implement a bootstrap estimator which is based on resampling the data until samples without monotone likelihood are encountered. Results from this analysis will be presented .