We apply various refinements of survival regression to assess the results of some basic specifications based on product life cycle theory for the case of a data set of the German automobile industry. The methods applied pay attention to biases in the coefficient estimates and the standard errors, the discrete nature of the duration data and the presence of unobserved heterogeneity. Robust estimation methods are also applied. We find that that the coefficient estimates and standard errors are not much affected by applying the refined estimators. The substantial results of a previous study with the same data are unchanged.