Different ways (methods) of running marginal effects

[Kanoknit Tanjan]

For multinomial probit in STATA, there are any other methods to run the marginal effect results for the continuous variable "MEMBERS" (which means number of household members). I simply add this variable in the syntax when running a marginal effect command. But I heard that it has sth like "at means", etc. What are the differences of the marginal effect result (if I just add the "MEMBERS" variable in the syntax or if I use "at mean")

. margins, dydx (MEMBERS)predict (pr outcome (1))

Average marginal effects Number of obs = 8102
Model VCE : OIM

Expression : Pr(Work_studycross==1), predict(pr outcome (1))
dy/dx w.r.t. : MEMBERS

------------------------------------------------------------------------------
| Delta-method
| dy/dx Std. Err. z P>|z| [95% Conf. Interval]
-------------+----------------------------------------------------------------
MEMBERS | -.0020091 .0019131 -1.05 0.294 -.0057587 .0017405
------------------------------------------------------------------------------


Thank you & regards,
Kanoknit
Social Researcher and PhD student, Thailand

[Richard Williams]

You might want to look at the section on "Adjusted Predictions and Marginal Effects" at

https://www3.nd.edu/~rwilliam/xsoc73994/index.html

Margins01 and Margins05 may be especially helpful. Margins01 provides an overview, including a discussion of asobserved (the default) and atmeans. (I think most people prefer asobserved instead of atmeans.) Margins05 talks about multiple outcome models including mprobit.

Incidentally unless you are using an old version of Stata (13 or earlier) you don't need to specify the outcome.

As a sidelight, your output will be easier to read if you use code tags. See the Statalist FAQ on asking questions effectively.

[Kanoknit Tanjan]

Many thanks for your advice.