> I think that I've found the reason. Once of the variables (m) > have to many missing values.... Yes, I think that is it. With the dataset UE_d.gdt, for example, the random effects model that you specify in your script cannot be estimated due to a lack of degrees of freedom: there are 4 independent variables, and only 4 usable panel units. However, gretl was not handling this error correctly. This should now be fixed: instead of a crash, you should get an informative error message.