Margins: default covariate treatment?

[km nicholas]

Hi folks,

I'm a graduate student and appreciate the wealth of resources on this forum. I've scoured the web and search function in Stata but I can't find an answer to this question: what is Stata's default approach to holding covariates constant in the margins function? I don't think it is at their means (hence why 'atmeans' is a specific command). I've seen language online indicating that covariates are "integrated across". Any insight would be much appreciated!

Khris

[Clyde Schechter]

Well, the default treatment of covariates in -margins- is not to hold them constant at all. Any covariate whose values are not constrained by an -at()- or -atmeans- option are simply left at their observed values in the data. The result is that, by default, effects estimated by -margins- are appropriately adjusted to the distribution of the observed values of the covariates.

If you want the estimation to constrain the covariates to specific values, you have to specify them in -at()- or by -atmeans-.

[km nicholas]

That makes perfect sense, thanks very much Clyde!

[Richard Williams]

The default is asobserved. For an intro to the margins command, you can see

https://www3.nd.edu/~rwilliam/stats3/Margins01.pdf

[Maxence Morlet]

If ever someone is interested, in addition to the great responses already provided, this paper is very relevant:

https://onlinelibrary.wiley.com/doi/...7.2012.00602.x.