Code structure for "Non-Gaussian Dynamic Bayesian Modelling of Panel Data", by Jurez & Steel

The code has three main programme types:

	n*.m	perform one update of the * parameter
	pdf*.m	calculate the (conditional) pdf of the * parameter
	the rest perform some necessary processes

main.m is the main routine.  To fit the model you need to call this file in matlab as

[MCMC ACC] = main(data, covs, burn, reps, thin)

where

MCMC is the output structure
ACC is an structure with the acceptance rates for the Metroplis steps

data is a T X M matrix, where missing data points are NaN values
covs is a p X T X M matrix where p is the number of covariates and where missing data points are NaN values
burn is the initial burn-in period
reps is the total number of MCMC repetitions
thin is the thinning factor.