Creating a variable to be used as pweight

[Oscar Ozfidan]

I need some input as to whether my understanding of pweight is correct.
Suppose there are two categories of observations with different chances of being included in the sample:
so the pweight for the first category would be n1/N, and for the second category it would be n2/N.
suppose n1 is smaller than n2, and therefore n1/N would be smaller than n2/N.
My confusion starts right at this point. Since pweight is supposed to be the inverse of the ni/N, do I need to make that calculation and
define the pweight as 1/(n1/N) for the first category and 1/(n2/N) for the second category, or do I need to leave it as (n1/N) for the
first category and (n2/N) for the second category? I guess what I am not sure about is whether Stata incorporates the inverse of ni/N automatically or it expects the user to
provide that calculation.

[Clyde Schechter]

The variable that you set as the pweight must already be the inverse of the probability of sampling. Designating it as a pweight does not cause Stata to change its values.

If you don't find what it says in the documentation about this clear, you could also have verified this by opening up any of the survey data sets from the StataCorp website (e.g. webuse nhanes2) and run -summarize- on the designated pweight. You would see that all of the values are >= 1, so they must be inverses of probabilities, as probabilities are always <= 1.