Adjusted Wald Test Interpretation

[Alina Faruk]

Apologies in advance for this extra rookie question.

Apparently I cannot use a t-test for survey data, and I will have to resort to Adjusted Wald test to test differences in means:

https://stats.idre.ucla.edu/stata/fa...h-survey-data/

I am using Method 3 from here.

I am having some confusion regarding the interpretation of the Adjusted Wald Test. Can someone please verify if I am thinking right from the following two examples:

So basically for this one:

Variable	Difference
Training	0.0001

Next one:

Variable	Difference
Training	-0.0057***

Am I doing this right?

[Carlo Lazzaro]

Alina:
in your first example, switching from male to female does not cause any evidence of a statistical significant effect on the regressand. The difference between the -constant- (that is, male in your case) and -female- coefficient is (after rounding) 0.0001, as you reported. Wald test mirrors exactly what the p-value in the regression outcome table tells you (.9328 vs .933; the difference is entrirerly due to a small rounding matter).
The opposite holds for your second example: switching from male to female does cause any evidence of a statistical significant effect on the regressand. The difference between the -constant- (that is, male in your case) and -female- coefficient is (after rounding)-0.0057, as you reported. Again, Wald test mirrors exactly what the p-value in the regression outcome table tells you (.0000 vs .000; the difference is entrirerly due to the number of informative digits after the separator.).

[Alina Faruk]

Alina:
in your first example, switching from male to female does not cause any evidence of a statistical significant effect on the regressand. The difference between the -constant- (that is, male in your case) and -female- coefficient is (after rounding) 0.0001, as you reported. Wald test mirrors exactly what the p-value in the regression outcome table tells you (.9328 vs .933; the difference is entrirerly due to a small rounding matter).
The opposite holds for your second example: switching from male to female does cause any evidence of a statistical significant effect on the regressand. The difference between the -constant- (that is, male in your case) and -female- coefficient is (after rounding)-0.0057, as you reported. Again, Wald test mirrors exactly what the p-value in the regression outcome table tells you (.0000 vs .000; the difference is entrirerly due to the number of informative digits after the separator.).

Thank you so much for the elaborate responses Mr Lazzaro!!!