nickseedorff/pmmhLong: Pseudo-marginal Metroplis Hastings for Efficient Multivariate Longitudinal Poisson Regression

Latent gaussian processes can be used to model multivariate count data by including available structural or functional distances between observations. Traditional MCMC approaches can be slow in this setting given the dimensionality of the unknown parameters and may have difficulty convering due the the flatness of the posterior distribution. Here we implement and efficient Pseudo-marginal Metroplis Hastings approach that uses an unbiased estimates of the marginal likelihood based on laplace approximations.

install.packages("remotes")
remotes::install_github("nickseedorff/pmmhLong")