Re: svy, subpop ( ): tab a b, obs Pearson failed to calculate Chi-square statistic

[Gebreselassie Tesfayi]

Greeting Stata Users,

I am using svy, subpop ( ): tab a b, pearson to generate chi-square statistic for a bivariate table. Variable 'a' has 3 categories and I wanted to drop one of the categories with n < 30. I created 0/1 variable, such that keepvalues=1 if inlist(a,1,2), 0 otherwise.
I used svy, subpop(keepvalues): tab a b, pearson to generate a chi-square statistic. However, instead of excluding the category the svy, supop (keepvalues) command populated the excluded category with zeros and failed to calculate the chi-squared.

Please advise.

The output after the run:
b
1 2
a 1 n11 n12
2 n21 n22
3 0 0

Table contains a zero in the marginals.
Statistics cannot be computed.

Pearson:
Uncorrected chi2(2) = .
Design-based F(., .) = . P = .

[Rich Goldstein]

-tab- is not one of the "estimation" commands allowed with -svy-; see

help svy_estimation

for what is allowed

maybe if you tell us more, a positive response can be provided

[Gebreselassie Tesfayi]

Thank you, Rich. But, tab is allowed with svy. Here is link: https://www.stata.com/manuals14/svys...latetwoway.pdf.
If I use "svy: tab a b if a < 3, pearson" it works, but the SE is not the same with using subpop() according to the link stata.com/support/faqs/statistics/subpopulation-estimation/
Looking forward your response.