Using -poisson- to generate Slope Index of Inequality(SII) estimates

[Otto Ruokolainen]

Dear Stata-listers,
(I posted an earlier version of this question also on Cross Validated but it was put on hold as off-topic)

I hope some of you would have some hints on how to proceed with my problem. I am using Stata SE15 to try to generate both Slope Index of Inequality (SII) and Relative Index of Inequality (RII) estimates for my binary dependent variable. I have managed to produce RII estimates but I am hoping to come up with a way to compute SII estimates.

I have a binary dependent variable y and two continuous variables (ridit and age) in my model. The estimates should be produced stratified by the study year and sex.

Earllier study has used the -glm- with the identity link (see code below) for the estimation of SII, but with this my model keeps iterating and does not convergence.

glm y ridit age if year==1 & sex==1, fam(bin) link(identity)

This is what (for example) the iteration 25 gives me:

Iteration 25: log likelihood = -1334.7175 (not concave)

Reading earlier studies, it may be for example, because of continuous variables in the model. As the literature points out, where some convergence problems occur, -poisson- could be used instead of -glm-. I have used both the -glm- and the -poisson- models for the RII estimates (see below) and they both work fine / give somewhat same results, but I haven´t found a way to produce the SII estimates with -poisson-.

glm y ridit age if year==1 & sex==1, fam(bin) link(log) nolog eform

poisson y ridit age if year==1 & sex==1, vce(robust) irr

Does anyone have any tips for me on how I could produce SII estimated with -poisson-? Any help would be greatly appreciated!

Best,
Otto