Is it possible to obtain relative risk rather than odds ratios using xtmelogit?

[Eduardo Torre]

Dear Statalist,

I am running my xtmelogit command without any problem but I would like to obtain RR rather than OR considering that for the distribution of my binary outcome the RR might be more appropriate. I know how to calculate RR using glm rather than xtmelogit but if I run the same model using glm it doesn't converge (which is strange considering that it does smoothly using xtmelogit). Any suggestion on how can I obtain RR using xtmelogit or any other command? Below my command:

xtmelogit preference i.sex i.rural i.year i.attainment i.income_q || group:, or
glm preference i.sex i.rural i.year i.attainment i.income_q group1 group2 group3 group4 group5, fam(b) link(log) eform

Thanks!

[Joseph Coveney]

You can Google convert odds ratio to relative risk and take a look at what comes up.

As far as other commands, meglm doesn't allow the combination of log link and binomial distribution family, but you could try the user-written command gllamm from SSC.

You appear to have only six groups . . . both melogit and gllamm are large sample estimation commands.