You can use the SOLVE statement in PROC MODEL to simulate multivariate data with a specified covariance matrix. Both normal and non-normal distributions are supported in PROC MODEL for simulation. The list of distributions supported for simulation in PROC MODEL includes normal, t, F, Cauchy, chisqaure, poisson, and uniform. These distributions are listed in the description of the ERRORMODEL statement in Syntax section in the MODEL procedure documentation.

You can also simulate multivariate data where different series follow different distributions.

The example code in the Full Code tab simulates bivariate data from normal, t, and uniform distributions using a specified covariance matrix. The last section of the example code simulates bivariate data in which one series is poisson-distributed and the other series is distributed as normal. In the example code, note that the WHERE clause following the OUT= option in the SOLVE statement removes the _REP_=0 results from the simulated data set. In releases prior to SAS 9.2, this is necessary when using the ERRORMODEL statement due to a defect documented in this problem note. Also, the example code should not be used prior to SAS 9.0 due to the defect documented in this problem note.