Also, see the methods in this note.  ... View more

PROC QLIM in SAS/ETS® can fit a bivariate probit model. ... View more

The question of sample size here is important.  As discussed in this note, the test for proportional odds is known to be liberal with small sample sizes. Your graphical assessment might be more important.  Also, as discussed in this note and in the "Details: Overdispersion: Rescaling the Covariance Matrix" section of the LOGISTIC documentation, the Pearson and deviance statistics require replication within the subpopulations in order to be valid. If there is suitable replication, then the similarity of the two statistics suggests they are providing a reasonable test of fit and their significance could be due to overdispersion or an incorrectly specified model.  You might want to try adding complexity to the model (interactions, quadratic terms, splines, etc.) as seems reasonable to try to achieve a correctly specified model. If these statistics are still significant, then you might have a problem with overdispersion.  The second note mentioned above discusses this.  ... View more

Since you describe a 2x2 table, you could avoid fitting a model altogether and simply use PROC FREQ and request an exact estimate of the odds ratio with the EXACT OR; statement. ... View more

I assume that you specified PARAM=ORDINAL because your response is ordinal.  You never need to specify the response variable in the CLASS statement, and I recommend that you don't.  Because you have PARAM=ORDINAL specified as a global option in the CLASS statement, it applies not only to the response but also your predictor.  That's why you don't get odds ratios by default (though you could probably add an ODDSRATIO statement to get it).  So, the following would be better code.  Note that if you need to model the probabilities of higher values of the response, use the DESCENDING response variable option as shown.   proc SURVEYLOGISTIC data=HRBM order=formatted; weight weight1; class risk(ref="no risk"); model Q169_2re(descending)=risk/CLODDS; run; ... View more

Try changing PARAM=GLM to PARAM=REF. ... View more

The offset variable is essentially a predictor variable in the model.  Any data that you want to score has to have values of the model predictors, so the offset has to be available. ... View more

Note that variable selection methods are available for zero-inflated models in PROC HPGENSELECT.   ... View more

You can model clustered, multinomial data using PROC GEE or using a CLUSTER statement in PROC SURVEYLOGISTIC. ... View more

The conditional model isn't possible, but you can model clustered, multinomial data using PROC GEE or using PROC SURVEYLOGISTIC with a CLUSTER statement.  Both allow you to fit the model without having to estimate cluster-specific parameters. ... View more

Currently, there is no SURVEY procedure for fitting count models using the Poisson or negative binomial distributions.  SURVEYREG will assume that the response is normally distributed.  ... View more

PROC GENMOD should not be used to analyze complex survey data. Only the SURVEY procedures (SURVEYFREQ, SURVEYLOGISTIC, etc.) can provide a proper analysis of survey sample data.  A variable specified in the WEIGHT statement in other procedures may produce correct parameter estimates, but their variances will not be correct.  Special variance estimators are needed in the analysis of survey data and only the SURVEY procedures have these estimators.  There is currently no SURVEY procedure available for fitting count models such as Poisson or negative binomial models.   See this note for more.  ... View more

See this note. ... View more