Hello,

you can see the link from stata help,

https://www.stata.com/statalist/archive/2005-06/msg00069.html

Also, the same question had been asked in RG plateforme and you find in the attached link the responses of certain professors:

https://www.researchgate.net/post/Why_in_multinomial_logistic_regression_STATA_does_not_produce_Odds_Ratio_rather_it_produces_Relative_Risk_Ratio

Best Wishes

Chellai Fatih

Thanks for your comment. I have read those links and comments. But, which made me more confused is that, if I perform multinomial regression in other software (SPSS, SAS) then, will they also provide RRR in lieu of OR?

But what's the problem when a software provides you OR rather than RRR or inverse?

Chellai Fatih

Obvious you know, there are a great difference between OR and RR.

Because, OR = ad/bc, and RR = (a/(a+b))/(c+(c+d))..

Yes this is so clear dear Asibul Islam Anik ,

The meaning of my question, is you interpret the estimation results accroding to the output ( OR or RRR).

But, just another idea about STATA, it consider the OR and RRR the same ( as Exp(beta)) (i.e) it puts OR as an approximation of RRR, despite certain biais of this approximation. (Camey et al, 2014)

Article Bias of multinomial logistic regression to estimate relative...

Furthermore, you can use the command mlogit, rrr after running the multinomial logit model or by specifying the rrr option when the full model is specified.

Best wishes

Chellai Fatih

Yes..That is the thing I wanted to know..that..Do STATA consider OR & RRR as same estimation?

Thanks a lot ...I got my answer now..