Ted:

The answer is simple. It is because SAS is modeling the category 0 of the dependent variable. In order to change the default, you need to add the DESCENDING option in the PROC LOGISTIC statement.

For more information of getting around the default setting of PROC LOGISTIC, you can read the article "If There Is A Will, There Is A Way: Getting Around Defaults of PROC LOGISTIC" in MWSUG '98 Proceeding. This article is also available in my Web site:

http://php.indiana.edu/~tso/articles/mwsug98.pdf

Harry So ------------------------------------------------------------------------------- Department of Educational Psychology | UITS Stat/Math Center School of Education | Indiana University - Bloomington Indiana University - Bloomington | 410 N. Park Avenue Bloomington, IN 47405 | Phone: (812)855-4724 -------------------------------------------------------------------------------

On Sat, 29 May 1999, Ted Floyd wrote:

> This has gotta be really basic, but I can't figure it out. > Can someone tell me why the following SAS program produces > parameter estimates and logarithmic odds ratios with the > WRONG signs? Thanks. Ted Floyd (tedfloyd@ableweb.net) > > Here goes: > > input H $ F; > cards; > 0 11 > 0 11 > 0 12 > 1 14 > 0 17 > 1 16 > 0 20 > 1 22 > 1 23 > 1 24 > PROC logistic; MODEL H = F; run; >