question for margins option(xb, fixedonly) in meqrlogit postestimation

[Jay Jang]

Your advice was very helpful, and thank you once again.

I am running the margins command after meqrlogit. (I am estimating using the melogit and meqrlogit commands.)
As you mentioned, the results were almost the same.

However, the margins result was slightly different with the margins, dydx(*) command.

1) the margins result after meqrlogit offer "Average marginal effects and Expression: Linear prediction, fixed portion, predict(xb)"
2) the margins result after melogit offer "Average marginal effects and Expression: Marginal predicted mean, predict()"

1) means 'xb calculates the linear prediction xβ based on the estimated fixed effects (coefficients) in the model. This is equivalent to fixing all random effects in the model to their theoretical (prior) mean value of 0.'

Can you explain how those two are different? and

What does it mean to use 'their theoretical (prior) mean value of 0', when estimating based on 'xb'?

Does it mean that the random effects are set to 0, and the variability between groups is not included in the model?

What does the result mean?

[Clyde Schechter]

Does it mean that the random effects are set to 0, and the variability between groups is not included in the model?

Yes, it means that the margins are calculated with random effects set to 0. The variability between groups is still included in the model--you've already estimated the model before you even get to the -margins- command, and -margins- makes no changes to those results. But the variability between groups is not reflected in the predictive margins results.