How to estimate population attributable risk (PAR) about multi-categorical variable

[Yusuke M]

Hello, statalists,

Now I am trying to estimate PAR about multi-categorical variable.

There is a categorical variable named x, which can have a number of 1, 2, 3, or 4.
Then, I want to estimate PAR; when all people x=1 moved to x=2.

More detail, there are 100 people,
Scenario 0: 20 people are x=1, 20 people are x=2, 30 people are x=3, and 30 people are x=4.

Then, I want to estimate PAR when;
Scenario 1: 0 people are x=1, 40 people are x=2, 30 people are x=3, and 30 people are x=4.
*Other covariates are same between Scenarios 0 and 1.

Maybe commands such as punaf is useful, but I cannot find the way to estimate in this situation....
Any answers and suggestions are helpful.

Regards,

Yusuke M