You may have problems with an outcome that cannot decrease. You may want to consider creating an outcome that represents the number of cigarettes smoked over a one year period.
You can place constraints using MODEL CONSTRAINT.
Jon Heron posted on Tuesday, March 09, 2010 - 5:20 pm
Thanks Linda, I see I need to read the manual more carefully - inequalities are mentioned on P557.
It seems quite common in tobacco growth modelling (a discipline I am new to) to have a measure which lacks some of the usual metric properties - e.g. a 5-level ordinal variable where the first category of never-smoker can never be revisited once you've had at least one cigarette.