What do you mean 'couldn't make it'? What was the error message? ... View more

I am not sure what you mean by a "logic model" (perhaps logistic?) nor am I sure what you mean by "flag variables". Furthermore, coding missing as 0 is often a mistake. Nevertheless, if you want to do so you can do it in an array (the following code is not tested): data new; set old; array misszero (*)  YOUR LIST OF VARIABLES HERE;   do i = 1 to dim(misszero);     if misszero(i) = . then misszero(i) = 0;   end; run; ... View more

Stepwise selection is not recommended. The p-values from the final model will be too low, the standard errors too small, the confidence intervals too narrow, the model too complex and the parameter estimates biased away from 0.   For a simple demonstration of this (albeit in GLM, not LOGISTIC, but the same thing applies) see my paper with David Cassell: Stopping Stepwise, here is the version we gave at NESUG a while back. For more proof, see Frank Harrell's book Regression Modelling Strategies (he uses R, but it's still a really good book :-)).  If you must use an automated method, I suggest LASSO or LARS in PROC GLMSELECT. Unfortunately, SAS does not yet have this for LOGISTIC, but I have found that getting several recommended models from GLMSELECT and then examining them in LOGISTIC works well. This may be even more so with an ordinal DV. Also, the last sentence in your paragraph starting "I have the problem...." seems to be cut off. As to your specific question, see the documentation, which states that SAS uses the chi-square statistic for each effect not in the model. ... View more

Rick's book on simulation is excellent (not surprising) and starts with the very basics. ... View more

Please show your code and say more about the variables. You ought to get at least one value, but some will be blank. See the examples in the documentation for PROC MIXED ... View more

In addition to Steve's points, all of which I agree with. Why is BMI categorized? Binning continuous variables causes problems. See Harrell. Have you corrected/noted the problems with BMI? Some of these are summarized in my post on Yahoo (very nontechnical) but there are additional problems when using it with children as good values change with age. Your predictor values seem highly likely to be colinear. You can check this with the COLLIN option in PROC REG (since colinearity is a function only of the predictors, the incorrectness of REG doesn't matter). ... View more

I agree with Steve's idea to use GLIMMIX or possibly NLMIXED, after checking out the residuals from the models generated by MIXED. GLIMMIX has a lot of distributions and with NLMIXED - well, there's almost nothing you can't do. There was an interesting paper at the most recent NESUG on fitting W shaped distributions. ... View more

Art - Exactly. I just wanted to make it clear in this thread that that was the case. ... View more

Yes indeed Art. But, as the link carefully notes, Michael Friendly's program assumes the original data are normally distributed. It might be interesting to experiment and see how far off results are with that macro by aggregating data, disaggregating it with the macro and then comparing results to t-tests (or whatever) on the original data. ... View more

Hi Gabriel You sometimes refer to Zero Inflated Binomial and sometimes Zero Inflated Negative Binomial. I think you mean the latter, and that led to Dorata_jarosz response (which is right for binomial, but not negative binomial). I don't see what a zero inflated binomial would be. If you upgrade to 9.3, you have access to PROC FMM which fits ZINB and ZIP models very easily. As to your statistical question: The Poisson model assumes the conditional variance is equal to the conditional mean. I have never found this assumption to be justified, although I suppose there must be cases where it is. I am not sure what you mean by "COUNTREG does not seem to handle the CLASS statement". The statement is part of the PROC (see the documentation) and I have used it in the past without problems. Can you post your code and log? ... View more

Thanks for referencing my paper, Art. VS, as Steve pointed out, Stepwise is a bad method even if you have all the data and computing time in the world. The p values are too low, the standard errors are too small, the parameters are biased away from 0... it's not good.  If you insist on an automatic method, Lasso or LAR is better; they are available in PROC GLMSELECT ... View more