topic Re: a question about p for trend and p for interaction in New SAS User

a question about p for trend and p for interaction

liiiiiiiib — Sat, 09 Oct 2021 07:25:56 GMT

hello everyone! I want to see whether the ComCat variable changes with sddsrvyr in the three category variable agegroup, that is, whether there is interaction between ComCat and sddsrvyr in different agegroups.

here is my code :
proc genmod data=diffcomcat;
class agegroup; weight wtmec12yr;
model comcat=agegroup sddsrvyr sddsrvyr*agegroup/ dist=p link =log type3 wald;
run;

But the line in agegroup=3 is 0 and there is no p-value, I don't understand it.
I want to ask whether the code is correct, and then how to understand the result.

Re: a question about p for trend and p for interaction

acordes — Sat, 09 Oct 2021 08:22:09 GMT

The 0 for the estimate of your last agegroup is because of your parameterization coding.
The last group is taken as a reference.
And there's no p-value for this level as the p-value reads as the hypothesis that level a differs from the reference level by x units.
You can overrule this setting some procedures by selecting other methods but the overall estimate and result isn't altered from that.
Read about the parameterization and design matrix in the Do Loop from Rick Wicklin.

Re: a question about p for trend and p for interaction

PaigeMiller — Sat, 09 Oct 2021 10:40:14 GMT

An explanation of why one category has a zero coefficient

https://communities.sas.com/t5/Statistical-Procedures/Interpreting-Multivariate-Linear-Regression-with-Categorical/m-p/591230#M28913