P\'olygamma Data Augmentation to address Non-conjugacy in the Bayesian Estimation of Mixed Multinomial Logit Models

摘要: The standard Gibbs sampler of Mixed Multinomial Logit (MMNL) models involves sampling from conditional densities utility parameters using Metropolis-Hastings (MH) algorithm due to unavailability conjugate prior for logit kernel. To address this non-conjugacy concern, we propose the application P\'olygamma data augmentation (PG-DA) technique MMNL estimation. posterior estimates augmented and default are similar two-alternative scenario (binary choice), but encounter empirical identification issues in case more alternatives ($J \geq 3$).