The mixed logit relaxes the restrictive IIA assumption, using simulated maximum likelihood estimate to allow for variation in individual taste for various attributes. This is accomplished by introducing a probability density function for attributes that are assumed to have heterogeneous value among respondents, h(β), enabling the estimation of the magnitude of this heterogeneity.