default control group in surveylogistic statement?

[knighsson]

Hello，

in my code below, I assign (diabetes=1 and type=2) as diagnosis=2, which represents Type 2 diabetes, and use proc surveylogistic  statement to investigate the relationship between the Type 2 diabetes and gender. 

Then, I assign(diabetes=1 and type=2) as diagnosis=1 (so diagnosis =1 represent type 2 diabetes now), and use the same proc surveylogistic statement to investigate the relationship. 

The inferential analysis results are different, which indicate that it matters whether diagnosis=1 or diagnosis =2 represents type 2 diabetes. So I want to ask should I use diagnosis=1 or diagnosis =2 represent type 2 diabetes. Which one would SAS consider as default control group?

Thank you!

DATA nh.A;
SET nh.A;
IF (diabetes=1 and type=2) then diagnosis=2; **** have T2D diabetes;
else diagnosis=1;
run;

*** variable diagnosis specified type of diabetes T2D;

*** SLR;
proc surveylogistic data=nh.A nomcar;
class gender / param=glm;
    strata SDMVSTRA;
    cluster SDMVPSU;
    weight glucwt4yr;
domain eligible;
model diagnosis=gender/ vadjust=none link=glogit;
lsmeans gender / cl oddsratio adjust =tukey;
run; 

[ballardw · Accepted solution]

The default behavior will assign the 'last' value as the "reference". You have several options to control that such as

Descending, sets the largest value as reference, Event= First or Last , reference='variable value' or keywords first and last.

The specific value doesn't matter as much as interpreting the results correctly using the reference value. Changing the value range with the same code may result in changes in direction but at core the result is the same.

I think you may be a bit cavalier in the way Type 1 diabetes is being treated though. You might consider using your Diabetes variable as the response and include Type in model as well as a class variable.

[knighsson]

Thank you!

This is very helpful!