09-06-2014 02:15 AM
. I was running SAS and I tried to calculate odd ratios of some variables of my dataset using proc logistic but I received a message that the model could not be fitted. After going into the SAS manuals and I was guided by SAS to set a new singular value to fit the model to my dataset. I checked the SAS manual and understood that the default singular value is 10-9 , then I changed the singular value for that value to 1 stepwise, I did this lots of times and in each time I used 10 times greater value (i.e. 10-9. 10-7, etc…). Finally SAS said that the model was converged but it dint give me any value in the tables, It only showed “.” Rather than values. How do I fix this please? Can I use Proc genmod rather than proc logistic to sort out the data with the odd ratios? Or it is better to use logistic?. many thanks
09-06-2014 08:40 AM
As far as I know proc genmod is designed for Possion Regssion . But you could try proc PROBIT which is almost like proc logistic , just have different Link Founction .
09-06-2014 09:47 AM
Seems like some linear combination of the explanatory variables are perfectly predicting the dependent variable. Nothing wrong with proc logistic and nothing will be improved using proc genmod or other procedures. Check your dummy variables seems like there is a complete association between dependent variable and some of the dummy variables.