I'm running into an odd problem while running a choice based phreg
I am modeling both categorical variables and one quantitative
variable. My dataset consists of a number of subjects, each making a
choice from a selection of choices (e.g. choosing one candy bar from a
selection of different candy bar brands).
The quantitative variable (e.g. price) will vary for each alternative
within the subject set.
My problem is that when modeling the quantitative variable, I am
getting 0 degress of freedom, and not returning a parameter estimate.
The categorical variables are estimated just fine, with each
alternative specific variable being assigned a parameter estimate that
identifies its utility against the comparison alternative. My guess as
to why this isn't working for the quantitative variable is because
there is no "comp" group, i.e. none of the alternatives within each
subject's choice set has a 0 value in the quantitative variable.
Can someone shed some light onto this?