topic Re: PROC Univariate and K-S test. in Statistical Procedures

PROC Univariate and K-S test.

godel_complete — Mon, 19 Oct 2020 18:49:16 GMT

Hello,

I have count data with associated dollar cost. I am using the PROC UNIVARIATE to determine, using the K-S, the Cramer-von Mises and the Anderson-Darling whether the Weibull distribution would fit the dollar amounts. When I use the PROC UNIVARIATE, however, I get a result that incorporates the latter 2 tests but does not include the K-S test. I am unsure why and would love to find out if its a mathematical or computational issue.

Re: PROC Univariate and K-S test.

PaigeMiller — Mon, 19 Oct 2020 18:53:26 GMT

Not all statistics are available in all situations. Please see this table: https://documentation.sas.com/?cdcId=pgmsascdc&cdcVersion=9.4_3.4&docsetId=procstat&docsetTarget=procstat_univariate_details53.htm&locale=en#procstat.univariate.tuniedf

Re: PROC Univariate and K-S test.

StatDave — Mon, 19 Oct 2020 18:58:36 GMT

The KS test is also available in PROC NPAR1WAY. If you want to use it to compare your data to a known distribution, you could generate data from that distribution using the distribution parameters estimated from your data, and then use the generated data along with your observed data to get the test.