# Dirichlet Process mixture model for spike sorting
Read me file for implementation of Dirichlet Process mixture model (DPM)
Nicole White, 2015. email: nm.white@qut.edu.au

The files contained in this folder allow for the estimation of a Multivariate Gaussian DPM using Markov chain Monte Carlo (MCMC).

The implementation of the sampler (slice sampler DPM.R) requires the following input/s:
	y : matrix of observations (PCs corresponding to each sampled spike).  
	Dimension is r (number of PCs/dimensions) x n (number of spikes/observations).  
	Code is currently set up to read in file as a .csv, using read.csv()

Values for each hyperparameter are specific as in the main text of the paper, in the list 'hypers'

For an initial number of clusters (K), the latent classification variable (z) is initialised using K-means

Information on the number of MCMC samples taken is contained in the list 'MCMC'.  Corresponding samples drawn at each (thinned) iteration are stored in 'MCMC.traces'.

NB: Implementation of main code file requires the following R packages to be installed: mcclust, MCMCpack, mvtnorm.