Dear all,
I am running a rank-ordered probit model using the cmp command. The dependent variable is travellers' preference for different time use type. Time use types are divided into five categories. Traveller answered the possibility of engaging in these activities with 1,2,3,4, and 5, with 1 indicating the most likely. There are ties in the ranking data. The cmp command is the following:
cmp (rest: timeuse1 = independent variables) (work: timeuse2 = independent variables) (nodo: timeuse3 = independent variables) (soc: timeuse4 = independent variables)(lei: timeuse5 = independent variables), ind((9 9 9 9 9)) tech(dfp) ghkd(1000,type(halton)) rev constr(5/8)
I am trying to obtain the probability of given outcome after cmp. For example, the predict command is the following:
* predict timeuse1=1,2,3,4,5, using (default) first equation
. predict pruse11, pr outcome(1)
. predict pruse12, pr outcome(2)
. predict pruse13, pr outcome(3)
. predict pruse14, pr outcome(4)
. predict pruse15, pr outcome(5)
No matter what I define as a given outcome, the calculated probability is the same, which seems to be counterintuitive.
Then, the predict command is the following:
* Predict all outcomes, all equations.
.predict prD*, pr
I get five probability values. And since there are 5 possible ranked values for each dependent variable, I think there should be 25 (=5×5) probability values.
Did I misunderstand the probability calculated by the rank-ordered probit model, or was the code wrong? If so, please correct my mistake
Thank you in advance.
0 Response to cmp command for rank-ordered probit model
Post a Comment