You are on the right track. It looks like you have successfully created an Oracle database connection but are missing the SCHEMA option on the LIBNAME. The hierarchy for a database server is as follows: Server, Database, Schema, Table. I'm assuming your ODBC DSN includes the server and database. Oracle SQL Developer should give you details of the required schema which may be displayed as a prefix on table names. ... View more

Thanks for your reply.  Do you mean not to transform but use beta distribution (with logit link)?   For the values in the iteration history, they seem like it should have converged at some point, but just keeps going.      ... View more

First you should create a new binary variable, sex_partners.  The 1-1 and 2-2 cases should receive one code (e.g. 1, "Y"), and the 3, 4, 5 should receive the alternate code (e.g. 0, "N").  This is your independent variable.  The other variables are nesting variables.   Then I would try:   proc glimmix data = mydatset  ASYCORR ASYCOV; class household pairs sex_partners; model disease = sex_partners /dist = bin  link = logit oddsratio solution; random intercept / subject=pairs(household); run; ... View more

Dates are stored as number of days. Datetime is stored as number of seconds.  To convert from DATETIME vales to DATE values all the DATEPART() function essentially doing is taking the remainder after integer division by (24 hours)*(60 minutes)*(60 seconds).     So when you take the datepart of values like 19,956 (which is 21st of  Auguest in 2014) then you get zero (which is the first day of 1960).   1220 data test; 1221 date='21AUG2014'd; 1222 datetime=date; 1223 put 'COMMA12. ' (date datetime) (= comma12.); 1224 put 'YYMMDD10. ' (date datetime) (= yymmdd10.); 1225 put 'DATETIME20. ' (date datetime) (= datetime20.); 1226 run; COMMA12. date=19,956 datetime=19,956 YYMMDD10. date=2014-08-21 datetime=2014-08-21 DATETIME20. date=01JAN1960:05:32:36 datetime=01JAN1960:05:32:36 What are you trying to do? ... View more

I’m not 100% sure, but you might be able to change the combination of ref= statements in the class statement. Class iv1(ref=‘0’) iv2(ref=‘1’); I would assume the two reference categories would end up as the reference combination for the interaction.   A similar option might be to actually create your interaction variable and define the reference group. ... View more

I've never personally done it, but I believe you can use proc glimmix with logistic assumptions.   proc glimmix data= survey; class dv; model dv = iv1 iv2 / dist=binomial link=logistic obsweight=weightvariable oddsratio solution; run; If binomial doesn't work, it might be bernoulli, which will give the same result.  I can't remember if glimmix allows both. ... View more