Dr. Iyer,

I have looked at your stata output which shows you have used Poisson Regression. So before your specific question can be answered, I must first point out that Poisson is not appropriate for a dichotomous variable such as episiotomy (Yes/No). Poisson is only appropriate for a count variable as the dependent variable (e.g. number of children; or number of episiotomies a women have had).

For your variable, a binary logistic regression (LR) will be more appropriate. Coincidently I wrote a paper in 2007 on episiotomy, using logistic regression. Although have used LR very frequently lately, I have attached my episiotomy paper for you. I hope it helps.

Dear Dr Gulzar,

Thanks a lot for sending me your paper. It is very very comprehensive.

I used a Poisson regression model as the prevalence of episiotomy was not rare (>10%). This is based on recent literature that recommends using poisson/cox regression with robust variance or log-binomial regression for cross-sectional studies with binary outcomes.

Links to the articles:

https://www.ncbi.nlm.nih.gov/pubmed/14567763

https://www.researchgate.net/publication/5293398_Approaches_for_estimating_prevalence_ratios

Link to my recent paper using Poisson regression: http://journals.plos.org/plosone/article?id=10.1371/journal.pone.0185739 Considering this backdrop I was wondering if you help me to accurately interpret interaction and margins.  

Thanks a lot,

Veena

I see that you are referring to a methods paper. They use Cox's Proportional Hazards Model and Poisson (rather than logistic). A lot of times nethods paper show better estimates for newly proposed methods than traditional methods. However, a general advice is that you don't want to repair if not broken. How about their conclusion " However, precautions are needed to avoid estimation problems in specific situations." If I was a reviewer of your paper I would have made you justify your choice, which will require a lot more work if you chose this new methods. Again, that is my take.....

Thanks alot Sir, for you very valuable reply. I will keep your point in mind!! :-)