Empirical identification in the mixed Logit model: analysing the effect of data richness