search

steve-the-bayesian / BOOM

A C++ library for Bayesian modeling, mainly through Markov chain Monte Carlo, but with a few other methods supported. BOOM = "Bayesian Object Oriented Modeling". It is also the sound your computer makes when it crashes.
GNU Lesser General Public License v2.1
36 stars 15 forks source link

Dynamic regression models need an option for a simpler prior. #23

Closed steve-the-bayesian closed 5 years ago

steve-the-bayesian commented 6 years ago

Right now the prior distribution assumes there are many dynamic regression coefficients and fits a hierarchical model to learn the variance of each. This can fail when there are only a few coefficients (and is almost certain to fail if there is only one).

The dynamic regression model should have an option to (a) choose a single Gamma(a,b) prior for all coefficient (inverse) variances, and (b) choose independent Gamma(a, b) priors, one for each variance.

steve-the-bayesian commented 6 years ago

Fix submitted to GitHub today. Will go live on CRAN on next push.