Sparse covariance estimation in logit mixture models

Summary This paper introduces a new data-driven methodology for estimating sparse covariance matrices of the random coefficients in logit mixture models. Researchers typically specify models under one two extreme assumptions: either an unrestricted full matrix (allowing correlations between all coefficients), or restricted diagonal no at all). Our objective is to find optimal subsets correlated which we estimate covariances. We propose estimator, called MISC (mixed integer covariance), that uses mixed-integer optimization (MIO) program block structure specification matrix, corresponding coefficients, any desired sparsity level using Markov Chain Monte Carlo (MCMC) posterior draws from matrix. The determined out-of-sample validation. demonstrate ability correctly recover true synthetic data. In empirical illustration stated preference survey on modes transportation, use obtain indicating how preferences attributes are related another.