Announcement

Collapse
No announcement yet.
X
  • Filter
  • Time
  • Show
Clear All
new posts

  • ipfraking weights

    Hi, I have a conceptual question about raking weights. I have a survey sample drawn from a known population. I used the population demographics sex, race, and age (I know these numbers but don't have the population dataset). SO using these population frequencies, I created raking weights for my sample. I used:

    ipfraking [pw=1], generate(rakedwgt) ctotal(PRData_age PRData_gender PRData_race)

    I don't have any sample weight variable and according to others' suggestion I used pw=1 in this case.

    It looks like this:

    Summary of the weight changes

    | Mean Std. dev. Min Max CV
    --------------+--------------------------------------------------
    Orig weights | 1 0 1 1 0
    Raked weights | 413.7 968.1 .8696 11144 2.34
    Adjust factor | 413.6847 0.8696 1.11e+04


    So after applying the raking weight to my sample, the proportions look similar to what I expected but the count is showing same as main population instead of being similar to the survey dataset frequencies.
    So when I created both weighted and unweighted tables for my variables, the unweighted N is what's in the actual survey dataset, while the weighted numbers are similar to whole population N. What should I do to adjust the raked weights?

Working...
X