Dear colleagues,

I have regressed an ordered ordinal outcome y = {1,2,3,4,5,6,8.9,10} on explanatory variables, so far using simply -oprobit y x1 x2 x3 x4 x5-. This has successfully told me the marginal impact of each of x1 through x5 on the value of y.

Now I would like to know how close each observation was to being placed one category lower. I think I would want to assess this with the distance between the latent score L_i and the next lower cutoff LC_i. When I look at the cutoff values, the distances between upper cutoff UC and lower cutoff LC differ across values of y, so presumably I should look for each observation i not just at (L_i - LC-i), but rather at (L_i - LC_i) / (UC_i - LC_i) to scale the distance by the total range of each latent score range?

If this reasoning makes sense, how do I best implement this in Stata? I have not so far figured out how to access the latent score L_i and lower cutoff LC_i for each observation i after running -oprobit-. It looks like I can also estimate an Ordered Probit model with -gsem- but would anyone understand how exactly I need to specify the command then?

Thanks so much!
PM