Generating and Applying Frequency Weights

[Evan Mark]

Hi all,

I am entirely new to Stata. I have a survey of 400 respondents that has oversampled females (it's 59-41 female to male). I need the ratio to be 52% female to 48% male. I am having difficulty finding ways to weight the data based on 52-48 female to male and then apply this weight throughout the data set. How do I apply weight to gender values in stata and apply across the survey?

I appreciate any help! I know this is a simple question, but I'm very new to this.

Thanks all!

Evan

[Weverthon Machado]

If I understand your question correctly, what you want is survey raking, a form of post-stratification. The package ipfraking might help you and it seems to handle complex designs (I haven't used it). But your case is as simple as it gets: matching the marginal proportions of a single variable to known population values.

This should do the trick, assuming you have a dummy variable for female:

* Safe way to get sample proportions
scalar nobs = _N
bysort female: gen sample_prop = _N/nobs

* Input population proportions
gen pop_prop = .
replace pop_prop = .52 if female
replace pop_prop = .48 if !female

* Generate weight
gen weight =  pop_prop/ sample_prop

* Compare:
table female
table female [pw=weight]