It's not always clear what people mean when they say "stepwise". In textbooks, this usually refers to some kind of completely algorithmic method in which the data analyst has no input beyond setting it in motion. That kind of stepwise selection is a bad idea. See the note on the Stata FAQ for reasons why.
http://www.stata.com/support/faqs/statistics/stepwise-regression-problems/If you mean that you want to control which variables get added (or removed) on each "step", that is usually called hierarchical regression. And IIRC, only the REGRESSION procedure (for linear regression) allows you to have multiple steps in one call to the procedure. If you are using some other procedure (e.g., PLUM, which you may be using with your ordered outcome variable), you will have to call the procedure for each step separately. One thing to watch for is that you use the same number of cases for each model -- i.e., when you add variables, you can lose cases due to missing data. One way to do that is by setting a filter for cases that have valid data for all variables included in the most complex model.
Oh yes, one more thing. Because you have to run each step as a separate call to PLUM (or whatever procedure you're using), you'll have to do the model comparisons yourself too. For models fit by MLE, you'll compare the -2LL values. The change in -2LL is approximately distributed as chi-square with df = the difference in the number of parameters between the 2 models being compared.
HTH.
torvon wrote
Hey there.
I have some correlated predictors (.2 to .3) and an ordered outcome variable. I want to use a regression.
What I'm looking for is basically stepwise regression to make sure to control for correlations between predictors, but for ordered variables.
I don't find any way to do this in SPSS.
Any help would be greatly appreciated.
Thanks
E
--
Bruce Weaver
bweaver@lakeheadu.ca
http://sites.google.com/a/lakeheadu.ca/bweaver/"When all else fails, RTFM."
PLEASE NOTE THE FOLLOWING:
1. My Hotmail account is not monitored regularly. To send me an e-mail, please use the address shown above.
2. The SPSSX Discussion forum on Nabble is no longer linked to the SPSSX-L listserv administered by UGA (
https://listserv.uga.edu/).
Locked
