Solved – R : using survey package to run t-test on sub population of weighted data set

rsurvey

I have a large set of weighted data. I have loaded it into a survey design and would now like to run t-tests on sub-populations.

example:

DF<-cbind(ID, WEIGHT, GENDER, INCOME)

     ID WEIGHT GENDER RELOCATE INCOME
[1,]  1   4380      1        1     35
[2,]  2   5000      1        1     20
[3,]  3      0      0        1     55
[4,]  4   5640      1        0     60
[5,]  5   6120      0        1     25

example.survey<-svydesign(ids=~0, data=DF, weights=WEIGHT)

I am able to call the mean income for the entire sample by:

svymean(INCOME, example.survey)

       mean    SE
[1,] 35.227 9.043

However, I want to compare the means for a subpopulation of this sample so that I can maintain the proper weights.

Can you confirm that this is the proper syntax to run a t-test comparing the mean INCOME based on GENDER for those who relocated (RELOCATE==1)?

svyttest(INCOME~GENDER+RELOCATE==1, example.survey)

data:  INCOME ~ GENDER + RELOCATE == 1
t = 0.9841, df = 2, p-value = 0.4288
alternative hypothesis: true difference in mean is not equal to 0
sample estimates:
difference in mean 
          14.78145

Best Answer

you want

svyttest( INCOME ~ GENDER , subset( example.survey , RELOCATE == 1 ) )
Related Question