Proc surveyreg - how to compare the differences in means for groupings

einstein — Wed, 08 Jun 2016 06:28:12 GMT

I am trying to compare the % of adults who are uninsured with hypertension in 2013 vs. 2014 for democratic states and similarly for republican states. See table below:

Hypertension
Democratic States	Democratic States	Republican States	Republican States
2013	2014	2013	2014
12.4%	10.0%	18.7%	14.0%

For this, I used the proc surveyreg procedure to get at the means and compare them to see if they were statistically significant. However, what I really want to do is compare the difference in means between 2013 and 2014 for democratic states compared to republican states to see if its statistically significant.

So using the example data above, is a 2 percentage point difference for democratic states statistically different from the 4% percentage point difference for republican states. Does anyone know how I can calculate that?

proc surveyreg data=one; 
strata strat_p; 
cluster psu_p; 
class srvy_yr dem_or_repub; 
model hypertension = srvy_yr*dem_or_repub/ vadjust=none; 
lsmeans srvy_yr dem_or_repub srvy_yr*dem_or_repub / diff; 
weight wtfa_sa; 
run;

Thanks!

Re: Proc surveyreg - how to compare the differences in means for groupings

ballardw — Wed, 08 Jun 2016 15:31:17 GMT

Why did you start with Surveyreg (other than you likely had a complex sample)?

Re: Proc surveyreg - how to compare the differences in means for groupings

einstein — Wed, 08 Jun 2016 16:49:13 GMT

Yes, that's correct its a complex survey design.

topic Re: Proc surveyreg - how to compare the differences in means for groupings in Statistical Procedures

Proc surveyreg - how to compare the differences in means for groupings

Re: Proc surveyreg - how to compare the differences in means for groupings

Re: Proc surveyreg - how to compare the differences in means for groupings