We conduct extensive Monte Carlo experiments on non-parametric estimations of duration models with unknown duration dependence and unknown mixing distribution for unobserved heterogeneity. We propose a full non-parametric maximum likelihood approach, based on time-varying lagged explanatory...