Mixed Logit

Mixed logit is a fully general statistical model for analyzing discrete choices, extending the standard logit model by letting the coefficients on the explanatory variables vary randomly across decision-makers, drawn from a chosen probability distribution such as the normal or lognormal, rather than remaining fixed for the whole population. This lets the model capture random variation in individual tastes, avoid the restrictive independence-of-irrelevant-alternatives property of standard logit, and account for correlation over time in repeated choices from panel data. Mixed logit can approximate, to any degree of accuracy, any true random-utility model of discrete choice, and it is widely used in transportation, marketing and consumer-behavior research.