Factor of Covariate

June 21, 2005 1:33 PM

I'm trying to run an ordinal regression in SPSS. Are my independent variables factors or covariates?

I just want the ordinal regression equivilant of taking a bunch of variables and throwing them in a regression. Variables have different levels of measurement. (some dichotomous, some continuous etc.). Which box do I throw them into?

posted by duck to computers & internet (6 comments total)

Typically, in GLMs, factors refer to categorical predictors and covariates refer to continuous predictors. I haven't used SPSS to do an ordinal regression, but I would imagine that it is the same here.
posted by naturesgreatestmiracle at 2:09 PM on June 21, 2005

Are you asking about the difference between factors and covariates in general? If so, independent variables are factors. Covariates are generally uncontrolled variables that are thought to affect the response - they are thought of as error, if not treated as covariates.
posted by jasper411 at 2:50 PM on June 21, 2005

Jasper411: But in OLS regression independent variables and control variables are mathematically identical, so there's no reason you would have to distinguish them for the computer.

Naturesgreatestmiracl: I assume that by categorical you mean nominal (as opposed to ordered categories)? What about dummy variables?
posted by duck at 4:50 PM on June 21, 2005

Okay, so it's a drag and drop thing. Looking at a sample dialog box here, it appears that it's making a categorical/continuous distinction.
posted by ROU_Xenophobe at 5:39 PM on June 21, 2005

I assume that by categorical you mean nominal (as opposed to ordered categories)?

Right. You can use the dummy coded variables (or effects coded or contrast coded) variables in the covariates box if you want to get the coeffcients for the particular categories.
posted by naturesgreatestmiracle at 7:57 PM on June 21, 2005

It is late and I am sleepy, but the "ordinal" bit refers to the dependant variable. I know that the question may have been answered already, but I am still trying to figure out what the question is. Do you want to look at using nominal independent variables with an ordinal dependant?
posted by cgk at 9:01 PM on June 21, 2005