Multinomial logit with non-exclusive dependent variable

[Katie F]

Hi -

I would like to run a multinomial logistic regression on panel data. However, my dependent variables are non-exclusive (one person can be in multiple categories at a single time point).

It was suggested that I could transform the data and weight the lines that use duplicate values at 0.5.

This would involve something like this --
Going from data that looks like this:
ID Timepoint 1 2 3
1 1 0 1 1
1 2 0 0 1
2 1 1 1 0
2 2 1 0 0
3 1 0 1 0
3 2 1 0 0

To this (I've marked the changes with **):
ID Timepoint 1 2 3
1 1 0 1 0 **
1 1 0 0 1 **
1 2 0 0 1
2 1 0 1 0 **
2 1 1 0 0 **
2 2 1 0 0
3 1 0 1 0
3 2 1 0 0

Any suggestions for how to go about this transformation in Stata? Is this a reasonable plan or would I do better to run separate models on the output or consider another alternative?

Thanks very much for your feedback!
Katie

[Joseph Coveney]

Here's a naive suggestion: why not make the outcomes mutually exclusive? Someone with Outcomes 1 and 2 (out of a possible three) would be assigned Outcome 4, another with Outcomes 2 and 3 would be assigned Outcome 5 . . .

[Katie F]

Absolutely -- it's a great point and normally would be an ideal solution, except the permutations of outcomes are many (>30), and the result doesn't mean all that much since I'm interested in the individual practices as opposed to the pairings. If you have any thoughts on other possibilities, do let me know!