Calculate difference in ratio/proportions between two survey samples using proc surveymeans

[fieldsa83]

I can do this for difference in weighted frequencies by making the previous month's employment variable negative (-1) and the current month's employment variable positive (+1), which produces the net sample-to-sample change in the variable "employed", as well as the CV. But I don't know how to do it for ratios (for example the total rate of employment).

data output.newcombined ;

set output.combined;

if employed=1;

if sampledate in (previousmonth) then changeemployed= -1;

if sampledate in (currentmonth)  then changeemployed=  1;

run;

proc surveymeans data=output.newcombined varmethod=brr clm nobs mean median CV  cvwt SUMWGT Sum CVSum CLSum  alpha=0.32;

    var changeemployed;

    weight finalwt ;

    repweights bw1-bw1000 ;

run ;

[ballardw]

I'm not really clear on what you are attempting to do. Are attempting to compare between to groups of records? That might mean you want a DOMAIN analysis to generate the summaries for the two groups and then examine the results.

Is your employed variable a 0/1 dichotomous to indicate a status? then perhaps that varaible should be the VAR and Sampledate or your changeemployed as DOMAIN variables.