Hello SAS experts   In a couple of earlier forums I have been discussing modelling out of pockets cost for members of a health plan. There are a large number of zeros in this very large dataset(about 80% of 5.5 million observations) and the non-zero out of pockets are highly right skewed(from a few dollars to $55,000).   So I've tried PROC HPFMM and PROC HPGENSLECT with a Tweedie ,Gamma or ZINB distributions, although these are not really count data. The zeros are not a latent class variable either, as they don't come from a different process to those members who get an out of pocket charge. So I don't think finite mixture models, hurdle models or zero inflated models really work here.   Theoretically the Tweedie distribution should work but interestingly just doing plain ol' OLS with PROC HPREG seems to work best with the lowest AIC (by a long way with all these models); looking at the raw residuals(attached) with the OLS model shows the bimodal distribution quite well. Is this really valid though given the highly non-normal distribution of this outcome variable?   Modelling the out of pocket as binary with PROC HPLOGISTIC works well but I was trying to get a model which predicts the actual out of pocket, hence the attempts with the continuous outcome.   Thoughts appreciated.   Regards   Chris   ... View more

Hello SAS experts I'm using PROC HPGENSELECT to model heavily zero-inflated health insurance "out of pocket"(OOP) costs to health plan members. I'm puzzled why I can generate predictions with 5 explanatory variables but not seven. I just get blanks in the predicted column using the second code: proc hpgenselect data = OOP_logistic_CD technique=congra maxiter=1000 gconv=1e-4; class sex(ref=first) ageGroup5yr(ref=last) char_prov(ref="Mersey-Lyell (TAS)") hospGroup(ref="Melbourne Endoscopy Group Pty Ltd") specialtyName(ref="Psychiatry")/param=ref; model memberOop= sex ageGroup5yr char_prov hospGroup specialtyName/dist=Tweedie link=log; code File='scoringparameters.txt'; run; data ScoringData; informat sex $1. ageGroup5yr $5. hospGroup $60. specialtyName $60. char_prov $60.; input sex= ageGroup5yr= hospGroup= specialtyName= char_prov=; datalines; sex=F ageGroup5yr=45-49 hospGroup=Ramsay Health Care specialtyName=ENT char_prov=Inner Sydney (NSW) run; data Scores; set ScoringData; %include 'scoringparameters.txt'; proc print data=Scores; run; Here is the code using 7 predictors which just gives the blanks in the "Scores" dataset: proc hpgenselect data = OOP_logistic_CD technique=congra maxiter=1000 gconv=1e-4; class itemCatLev1Med(ref=last) sex(ref=first) ageGroup5yr(ref=last) char_prov(ref="Mersey-Lyell (TAS)") hospGroup(ref="Melbourne Endoscopy Group Pty Ltd") hospType(ref="Public" ) specialtyName(ref="Psychiatry")/param=ref; model memberOop= itemCatLev1Med sex ageGroup5yr char_prov hospGroup hospType specialtyName/dist=Tweedie link=log; code File='scoringparameters.txt'; run; data ScoringData; informat itemCatLev1Med $40. sex $1. ageGroup5yr $5. char_prov $60. hospGroup $60. hospType $40. specialtyName $60.; input itemCatLev1Med= sex= ageGroup5yr= char_prov= hospGroup= hospType= specialtyName=; datalines; itemCatLev1Med=Surgical Operations sex=F ageGroup5yr=45-49 hospGroup=Ramsay Health Care hospType=Private specialtyName=ENT char_prov=Inner Sydney (NSW) run; data Scores; set ScoringData; %include 'scoringparameters.txt'; proc print data=Scores; run; I suspect this is due to some collinearity although it doesn't seem that this should occur. Thanks for any advice. Chris ... View more