topic Wald test for whether the two coefficients are significantly different in SAS Programming

Wald test for whether the two coefficients are significantly different

ADou — Thu, 09 May 2019 19:49:36 GMT

Hi all,

I have two sub-samples, I also run regressions for the two samples separately. For sample1: y=x1+x2; for sample2 y=x1+x2. Now I want to test whether the two coefficients of x1 are significantly different? Should I use Wald test and how to realize it? Thanks in advance!

Re: Wald test for whether the two coefficients are significantly different

PaigeMiller — Thu, 09 May 2019 20:09:42 GMT

You create a class variable that indicates subpopulation (if you don't have one already) then fit this model

model y = subpopulation x1 x2 subpopulation*x1 subpopulation*x2;

If the term for subpopulation*x1 is significant, then this indicates there are different slopes for x1 for each subpopulation.

Re: Wald test for whether the two coefficients are significantly different

ADou — Thu, 09 May 2019 20:15:08 GMT

Thank you! Is it Wald test? Can I get wald Chi-square in this way?

Re: Wald test for whether the two coefficients are significantly different

PGStats — Fri, 10 May 2019 03:16:44 GMT

In proc GLM, ask for

model y = subpopulation x1 x2 subpopulation*x1 subpopulation*x2 / solution;

to see the estimates, their standard errors and the t-statistic tests equivalent to Wald tests.

Re: Wald test for whether the two coefficients are significantly different

PaigeMiller — Fri, 10 May 2019 10:50:02 GMT

@ADou wrote:

Thank you! Is it Wald test? Can I get wald Chi-square in this way?

That depends, are you using PROC LOGISTIC because you have binary Y-values?