topic Re: Proc model wald-test in Statistical Procedures

Proc model wald-test

sasphd — Wed, 01 Oct 2014 17:10:25 GMT

I want to test the equality of two parameters (a and b) in proc model. I write

test a=b

in the output I find type of test=wald , the statistic and ProbChiSq

Can someone help me to interpret the results?

Re: Proc model wald-test

ballardw — Wed, 01 Oct 2014 19:37:02 GMT

It is kind of hard to interpret something sight unseen. Try posting the results you are getting.

Re: Proc model wald-test

lvm — Wed, 01 Oct 2014 21:14:20 GMT

A wald test statistic is the estimate of the parameter (or contrast) divided by the estimate of the standard error of the parameter estimate (or of the the estimated contrast). This has a standard normal or t distribution under H0, depending on the circumstances. Square this and you get a chi-squared distribution. If the test is multidimensional (say, for two contrasts simultaneously), you can also calculate this using the vector of parameter estimates and the variance-covariance matrix of the estimated contrasts. Once again, a chi-squared distribution under H0.

Re: Proc model wald-test

sasphd — Mon, 10 Nov 2014 15:07:41 GMT

GMM Esti	Test0	Wald	42.57	<.0001	a=b

this is the result.

Can I conclude that a is different from b?

Re: Proc model wald-test

ets_kps — Mon, 10 Nov 2014 15:50:13 GMT

There is strong evidence that a is not equal to b based on what you have posted. AKA "A and B are different"

Now, I see that you are estimating these parameters with GMM so this might be some sort of complicated system. For that reason I cannot make any comment on the usefulness or appropriateness of this test statistic as I don't have much context.

Good luck-Ken