06-24-2016 05:04 PM
I am using data which has a complex survey design to get at number of uninsured for certain populations.
I tested to see if the percent of uninsured in one year was statistically different from the next year by using proc surveyreg and then using lsmeans with the diff option.
But does anyone know how do I test if the number (not proportion) of uninsured in one year is statistically different from another year. Would I need to use proc surveyfreq to do that?
proc surveyreg data=one; strata strat_p /nocollapse; cluster psu_p; class survey_year; model stroke = survey_year/ vadjust=none;
lsmeans survey_year / diff;