@Tendai wrote:
How do I replicate this SAS SUDAAN code
proc descript data= two filetype=SAS design=WR;
nest _ststr _psu/missunit;
weight _statewt;
subgroup cancer AGEp ;
levels 2 5;
var sex;
catlevel 1;
tables cancer;
stdvar agep;
stdwgt 0.42026284 0.20743092 0.20203482 0.08896694 0.08130448;
print nsum wsum percent sepercent lowpct uppct/wsumfmt=f6.0 totalfmt=f10.0 style=NCHS;
rtitle "age adjusted :";
output nsum wsum total percent sepercent lowpct uppct/wsumfmt=f13.0 totalfmt=f13.0
percentfmt=f12.1 sepercentfmt=f13.0 lowpctfmt=f13.0 uppctfmt=f13.0
filename=cancer3 filetype=sas replace;
run;
I do not have SAS SUDAAN and I would like to produce the same results that come out of SAS SUDAAN with SAS. The data has a weight "statewt" and its also age adjusted to 2000 population.
What SAS procedure do I use to get the same results that proc descript produces. Also what test would I use to test if sex is significant in the model.
Thank you,
For what it's worth, I don't think your SUDAAN code is doing what you think it's doing.
Your code asks SUDAAN to generate the proportion of records where sex=1 (var sex; catlevel 1) by levels of the cancer variable (tables cancer), standardized by age. I would think you'd want to generate the proportion of records with cancer (var cancer), by sex (tables sex), standardized by age.
You can see an example in creating age-standardized estimates with SAS proc Surveyreg here:
https://www.cdc.gov/nchs/tutorials/NHANES/NHANESAnalyses/AgeStandardization/Task1b_SAS92.htm
The examples look like they will produce point estimates and standard errors but I don't see a way to statistically compare standardized estimates between groups, such as sexes. You'd have to be able to construct some combination of model parameter estimates that corresponded to that test; that you'd then specify on an ESTIMATE statement.
