Dan Dalthorp
dalthorp@science.oregonstate.edu
Correlated count-valued random variables arise frequently in basic and applied ecology, social science, epidemiology, economics, and other disciplines in which counts are sampled through space or time. Statistical methods for analyzing correlated count data tend to be quite complex, and new analytical techniques are rapidly being developed. Simulation can be a valuable aid directly in statistical analysis and indirectly in the evaluation of new statistical techniques. Unfortunately, the repertoire of techniques for simulating correlated counts has been thin, and software for actually doing the simulations has been scarce. This webpage offers a partial remedy via a suite of Matlab functions to perform a variety of such simulations. Included are programs for generating the following types of correlated discrete random variables:
Given vectors of target means and variances of random variables to simulate and a matrix defining the correlations between them, the random variables can be generated using the Matlab utilities provided in the package discsim 2.2, which includes a User's Guide to the software. Brief descriptions of the simulation techniques are given here.