Weighted chi2 Tests

[Robyn Powell]

Hi,

I am a novice Stata user looking for help. I am using the Fragile Families and Child Wellbeing dataset. I have set my values to national weights: svyset [pweight=M1NATWT], jkrw(M1NATWT_REP*, multiplier(1)) vce(jack) mse


I am working on descriptive stats and have been asked to provide chi2 (or t-tests depending on variable).

I run a two-way and get this output:


. svy: tab CM1ETHRACE MOMID
(running tabulate on estimation sample)

Number of strata = 1 Number of obs = 1558
Population size = 466317.31
Replications = 33
Design df = 32


Mother
race
(baseline
own Does mom have ID?
report) no yes Total

1 white, .432 .111 .379
2 black, .241 .462 .278
3 hispan .245 .402 .271
4 other .0813 .0253 .072

Total 1 1 1

Key: column proportions

Pearson:
Uncorrected chi2(3) = 125.3162
Design-based F(2.01, 64.24) = 3.9822 P = 0.0233

My question is, how do I run chi2 tests on weighted values?

Thanks in advance!

[Steve Samuels]

Please see FAQ section 12 for how to format commands and results with CODE delimiters.

To answer your question, doing the svy: tab analysis automatically takes the weights into account. While the results quote a chi-square statistic, ignore it. The design-based F is based on that statistic and gives a more accurate p-value. So, for your example, p = 0.0233