Average Marginal Effects of all covariates

[Jason Cruso]

Hello everyone,

I am working on the following model (Random Effects Probit Model).

In the below model:
job_change = binary variable (0/1)
logsalary, education and tenure: continuous variables.

xtset  empid year
xtprobit job_change logsalary age education tenure, vce(robust)
margins, dydx(*)

The above looks OK to me as it estimates the average marginal effects of each of the covariates.

However, I also need to estimate the average marginal effects of all the covariates. i.e. the mean probability (Pr(y=1 | x̄))

I have tried the following model, but I wonder if this is correct since this would estimate the predicted probability.

xtset  empid year
xtprobit job_change logsalary age education tenure, vce(robust)
margins

I would really appreciate some help here.

Thanks a lot in advance.

J

[Felix Bittmann]

I understand that you want the following:

...
margins, atmeans

[Jason Cruso]

Thanks, Felix.
But I don't think so.

As per my understanding, the command that you suggest will give the adjusted predictions keeping all covariates constant.

margins, atmeans

Maybe someone can correct if I am wrong?

Thanks a lot,

J

[Felix Bittmann]

You wrote that you want "Pr(y=1 | x̄))", which I read as the probability to have received the treatment when all covariates are held constant at their respective mean. Maybe you can clarify what else you have in mind. In theory you can also combine these two commands into

margins, dydx(x) atmeans

[Jason Cruso]

Thanks, Felix for the clarification. You are right. I now understand this better.

J