converting OR into RR

[Pakeezah Saadat]

Hello,

I am running a propensity score analysis where imbalanced covariates are adjusted for using a logistic regression model but obviously, I get OR with log reg. Is there a way to convert it into RR or calculate it directly as RR?
I have come across marings and adjrr but for both of these, I need to run the logistic regression model first which cannot be done as I follow it up with psmatch2 command:

logistic trt var1 var2 var3
predict pscore
sum pscore

set seed 1000
generate x=runiform()
sort x

psmatch2 trt, out(var1 var2) pscore(pscore) caliper(.001) noreplacement descending common logit

So, as you can see, I cannot fit the margins or adjrr command before psmatch2 at all. Any tips / advice?

[Joseph Coveney]

So, as you can see, I cannot fit the margins or adjrr command before psmatch2 at all.

Sorry, I don't see. Why can't you issue a margins command just as you have a summarize command?

Is it because there something hidden, such as a return scalar from the latter that you need to avoid overwriting in order for it to be available for psmatch2? If so, then first margins and then summarize.

[Pakeezah Saadat]

I cannot use margins first because I need to adjust for imbalanced predictors which can only be known after I have run psmatch2. I cannot know how imbalanced my matching is until I actually do psmatch2.

[Leonardo Guizzetti]

On the topic of converting OR to RR, there is a nice discussion about it in an open-access BMJ article by RL Grant. The reference is Converting an odds ratio to a range of plausible relative risks for better communication of research findings. BMJ 2014; 348; DOI: 0.1136/bmj.f7450

The conversion from OR to RR is straightforward once the baseline risk (or odds) is known, but choosing the baseline risk may not necessarily be straightforward within your study.

[Pakeezah Saadat]

Thank you, I will look into the paper.