We propose a method to generate flexible mixture distributions that are useful for estimating models such as the mixed logit model using simulation. The method is easy to implement, yet it can approximate essentially any mixture distribution. We test it with good results in a simulation study...