Margins after (dynamic) panel probit

[Abdan Syakura]

Dear all,

I am trying to estimate the following (dynamic) panel probit model and then its marginal effect:

xtprobit y l.y x, re vce(robust)
margins, dydx(*) pr(pr)

I however received the following error message for the margins: "prediction is a function of possibly stochastic quantities other than e(b)". When I re-estimate the model without the lagged dependent variable, i.e. only

 xtprobit y x, re vce(robust)

, I did not have the error message.

Could you please advise how I can estimates the marginal effect for the first model? More fundamentally, am I wrong using xtprobit command for model with lagged dependent variable (dynamic)? Thank you.

Best regards,