Heckman selection-t model: Parameter estimation via the EM-algorithm

The Heckman selection model is perhaps the most popular econometric in analysis of data with sample selection. analyses this are based on normality assumption for error terms, however, some applications, distribution term departs significantly from normality, instance, presence heavy tails and/or atypical observation. In paper, we explore selection-t where random errors follow a bivariate Student’s-t distribution. We develop an analytically tractable and efficient EM-type algorithm iteratively computing maximum likelihood estimates parameters, standard as by-product. has closed-form expressions at E-step, that rely formulas mean variance truncated distributions. Simulation studies show vulnerability selection-normal model, well robustness aspects model. Two real examples analyzed, illustrating usefulness proposed methods. algorithms methods implemented new R package HeckmanEM .