| Date: | Thu, 19 Jun 2003 11:07:51 -0500 |
|---|
| Reply-To: | Robert Saunders <robert.c.saunders@VANDERBILT.EDU> |
|---|
| Sender: | "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU> |
|---|
| From: | Robert Saunders <robert.c.saunders@VANDERBILT.EDU> |
|---|
| Subject: | Marginal Effects in PROC LOGISTIC |
|---|
|
| Content-Type: | text/plain; charset="us-ascii"; format=flowed |
|---|
Hi,
Do I have to calculate "by hand" marginal effects (in terms of
probabilities) from PROC LOGISTIC? When I say "by hand" of course I mean,
"Program a solution with SAS"?
Does anyone have examples where they've done it? I've searched the
archives and couldn't find an example (if there are some, I couldn't find
them among too many false-positive hits).
I'm thinking this would roughly be my strategy to do it myself:
1. Make a table of the mean values of the RHS variables;
2. Run my PROC LOGISTIC step and get the OUTEST table;
3. Merge the two tables in such a way that I have my parameter estimates
paired one-to-one to my avg values of the RHS variables;
4. Calculate the sum of the XB X times (the number of RHS vars) to get the
effect of one-unit changes in each X var.
Am I way off base or making this unduly complicated? Or does one of the
other PROCs do this?
r
################################
Robert C. Saunders, M.P.P.
Box 90 GPC, Vanderbilt University
Nashville, Tennessee 37203
T: 615.322.8284
F: 615.343.2661
| 
