RSGHB: Hierarchical Bayesian Estimation Functions for Discrete Choice Models

Functions for estimating models using a flexible Hierarchical Bayesian framework, allowing direct user specification of likelihood functions. The package supports discrete choice models like Multinomial Logit and Mixed Logit, as well as ordered response models, with parameters specified as fixed or random across several distributions. It is based on Kenneth Train's original Gauss and Matlab code, rewritten and optimized for R with added flexibility.

Use Cases

• Estimate Mixed Logit models for consumer choice analysis based on the supported family of discrete choice models.
• Fit ordered probit or logit models for survey response data based on the described ordered response model support.
• Specify parameters with non-normal distributions like the Johnson SB for hierarchical modeling based on the listed parameter distribution options.
• Compare fixed and random parameter effects in a Bayesian framework based on the flexibility in parameter specification.

Strengths

• Provides a flexible framework allowing direct user specification of the likelihood function, not predetermined structures.
• Supports a range of parameter distributions including normal, log-normal, censored normal, and Johnson SB.
• Code is optimized for R, building upon and extending Kenneth Train's established Gauss and Matlab implementations.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.
• Last update date is unknown; freshness unverified.

Provenance

Source

Jeff Dumont, based on Kenneth Train's original Gauss and Matlab code from UC Berkeley.

Collection Method

Software package functions for statistical estimation.