Good question. I put more effort in discussions than referee reports. Is this only me?

Interesting question. I haven't seen anything like this. But would that not be a red flag in the sense that both OR and PS are misspecified -- and hence DR, too. Because if either OR or PS are correct, DR should be pretty close to one of them (in large samples).

What do you mean by different? Different sign and significant? Stat. diff but same sign?

Well then you were the lucky one...

Stranded in Dallas

Can happen if estimated with instrumental variables...

Ganz stark! Glückwunsch.

I see. Thanks for the link. I have not yet started to think about automation, but that's a good starting point.

Is that what you are looking for?

par(mar=c(3,3,1,1)+0.5, mgp=c(1.5,.5,0), tck=-.02)

I got this from recent post by Andrew Gelman and I like it.