Set of tools to fit a linear multiple or semi-parametric regression models with the possibility of non-informative random right-censoring. Under this setup, the localization parameter of the response variable distribution is modeled by using linear multiple regression or semi-parametric functions, whose non-parametric components may be approximated by natural cubic spline or P-splines. The supported distribution for the model error is a generalized log-gamma distribution which includes the generalized extreme value and standard normal distributions as important special cases. Also, some numerical and graphical devices for diagnostic of the fitted models are offered.