Negative individual predicted probabilities were identified in the final model fit

[Demographer]

Hello,

I want to perform a generalized ordered logit regression to predict education in 3 levels (edu3) with a set of variables. I tried this:

proc logistic data=edu.educationClean;
class  Langue(ref='0') cntr_regr(ref=first) genstat3(ref='0') eduM3(ref='2') religion(ref='0') /param=ref;
model edu3(descending)= cntr_regr lncohort cntr_regr*lncohort Langue genstat3 religion eduM3/unequalslopes maxiter=500 rsquare;
weight pond2 / norm;
by EU15 sex;
run;

However, I got this warning message:

WARNING: Negative individual predicted probabilities were identified in the final model fit.  You
         may want to modify your UNEQUALSLOPES specification.
WARNING: The LOGISTIC procedure continues in spite of the above warning. Results shown are based
         on the last maximum likelihood iteration. Validity of the model fit is questionable.

When I look at the results, they look consistent. What does that mean exactly? How can I fix the problem? I also ran a multinomial logit model on the same dataset and there were no message like this.

Thank you.

[Ksharp]

1) Remove   UNEQUALSLOPES 

2) variable EDU3 has value like 1 2 3?

[Demographer]

1) I need this statement, because I want 2 set of parameters for all variables.

2) Values are 0-1-2

[Ksharp]

That is weird. LOG said you got negative predicted prob, that doesn't make any sense .

Try remove "

weight pond2 / norm;  

or use 

freq pond2;

[Demographer]

Same message.