1

Let's say I have a survey question on political party affiliation - Democrat, Republican, Independent, Other. I want to test whether the weighted proportions are equal.

I cannot use Base R functions because they don't take account of sampling weights in the calculation of variance.

I have tried svychisq() but that only works for testing for association between two variables.

user242
  • 11
  • 2
  • how about `stats::prop.trend.test()`? – David Z Apr 20 '21 at 00:25
  • 1
    @DavidZ Thanks for the comment, but that will not work for several reasons. I have never used prop.trend.test() before, but it looks like it is testing for a trend, e.g., increasing or decreasing. My data is not ordinal, so there doesn't have to be a trend for there to be differences between proportions. Also, the test does not account for the sampling design. I believe only R Survey accounts for the sampling design when performing tests. – user242 Apr 20 '21 at 00:48

0 Answers0