topic Re: regression results produced a categorical variable with no parameter estimate outputs - why is t in Statistical Procedures

regression results produced a categorical variable with no parameter estimate outputs - why is that?

sams8765 — Thu, 04 Nov 2021 18:52:17 GMT

Hello - This is not a coding question but an interpretation question. I ran a regression model in SAS and the model results look good except I notice that one of my categorical predictor variables is missing from the parameter estimates. What does this mean? How do I interpret these results?

Re: regression results produced a categorical variable with no parameter estimate outputs - why is t

Reeza — Thu, 04 Nov 2021 19:08:26 GMT

Is there a note in the log?
Is it a level of a categorical variable or all levels for that categorical variable?

Re: regression results produced a categorical variable with no parameter estimate outputs - why is t

sams8765 — Thu, 04 Nov 2021 19:21:32 GMT

So it was just one level of my categorical variable. So overall, the categorical variable had 21 levels but only 20 of them showed up in the parameter estimates. I'm curious as to why just one of the levels of the categorical variable is missing?

There also was no notes/errors in the log that would indicate anything.

Re: regression results produced a categorical variable with no parameter estimate outputs - why is t

Reeza — Thu, 04 Nov 2021 19:25:38 GMT

How did you parameterize the variable? If you used reference coding that is expected, the estimate for that level, your reference level, get combined with the intercept.

Re: regression results produced a categorical variable with no parameter estimate outputs - why is t

PaigeMiller — Thu, 04 Nov 2021 19:43:30 GMT

@sams8765 wrote:

So it was just one level of my categorical variable. So overall, the categorical variable had 21 levels but only 20 of them showed up in the parameter estimates. I'm curious as to why just one of the levels of the categorical variable is missing?

There also was no notes/errors in the log that would indicate anything.

Did the 21st level show up with a zero coefficient, or is it not present in the output? Show us what you see.

Are two of the levels identical in the first 32 characters (even if they are different in the 33rd or later characters)? What is the name of the level that is missing?

Re: regression results produced a categorical variable with no parameter estimate outputs - why is t

ballardw — Thu, 04 Nov 2021 20:06:04 GMT

Since there many regression procedures perhaps you can share code.

That way we know which options you used that might affect output. Then also sharing the specific output you are looking at and which parameter.

Re: regression results produced a categorical variable with no parameter estimate outputs - why is t

sams8765 — Fri, 05 Nov 2021 13:04:19 GMT

Here is the regression code I am using. Please note that it's embedded within a macro. I am only including the actual proc hpgenselect statement.

Here is a screenshot of the output. The predictor variables used as input was A*B (so the interaction of these 2 variables where A has 2 levels and B has 23 levels (i originally send 21)). The 5.1 relates to the intercept and then the first set of 1 through th relate to A level 1. You'll notice the A level 2 starts with 02 so the 01 is missing. This is what I'm trying to solve. Why would the (A level2)-01 be missing?

Re: regression results produced a categorical variable with no parameter estimate outputs - why is t

PaigeMiller — Fri, 05 Nov 2021 14:39:16 GMT

Oh come on, why be so stingy with information? Show us a screen capture of the output that is not chopped off so we can read the labels and columns headings. And show us the values of your macro variables &class and &target and &predictors so we might understand what the model is you are creating.

Re: regression results produced a categorical variable with no parameter estimate outputs - why is t

Reeza — Fri, 05 Nov 2021 16:02:18 GMT

How many observations do you have in your data set?