What distributions did you have in mind? If you can compute the CDF (i.e., with one of SAS's cdf functions or approximate the integral numerically or some other formula), you can compute GOF with a few SORTs and DATA steps.

Robin High University of Oregon

-----Original Message----- From: SAS(r) Discussion [mailto:SAS-L@LISTSERV.UGA.EDU] On Behalf Of OR Stats Sent: Thursday, January 24, 2008 12:34 PM To: SAS-L@LISTSERV.UGA.EDU Subject: Fitting and Testing Goodness of Fit to non-Standard Distn in SAS

Dear All:

I would like to fit and test the goodness of fit of my empirical data to distributions. Unfortunately, these two distn's are not pre-coded in SAS. So while PROC UNIVARIATE has nice features for fitting and testing the fit of our data, it appears to work only if we are interested in fitting our data to Beta, Exponential, Gamma, Lognormal, Normal, or Weibull. What is the most efficient way of fitting and testing goodness of fit of our data if the distn that we assume is not one of these? Is Proc Model the best way to do this in SAS?

While one of my distn's has closed form, the density of my other assumption does not. So if I would like to fit my data to a non-closed form density function, and again test for goodness of fit, how can I best do this in SAS?

Thank so much, ORstats