marginal_effects() for nonlinear link functions in hierarchical models

[Dimitris Rizopoulos]

I have a question for the predictions / fitted values produced by marginal_effects() for hierarchical models with random effects and nonlinear link functions (e.g., mixed effects logistic regression): Namely, in the help page of the function it is mentioned that when re_formula = NA (default), then marginal predictions are calculated. Does this mean that only the fixed-effects part of the model is used? For nonlinear link functions this wouldn’t correspond to marginal predictions, right?

Thanks for any pointers!

[Paul Buerkner]

Yes indeed, only the fixed effects are used, which in non-linear models is not identical to marginal predictions. Unfortunately, the latter are far more difficult to compute...

In retrospect, the name "marginal_effects" was suboptimal. See also my comment in the details section of ?marginal_effects

2018-03-01 23:29 GMT+01:00 Dimitris Rizopoulos <dimitris....@gmail.com>:

I have a question for the predictions / fitted values produced by marginal_effects() for hierarchical models with random effects and nonlinear link functions (e.g., mixed effects logistic regression): Namely, in the help page of the function it is mentioned that when re_formula = NA (default), then marginal predictions are calculated. Does this mean that only the fixed-effects part of the model is used? For nonlinear link functions this wouldn’t correspond to marginal predictions, right?

Thanks for any pointers!

--
You received this message because you are subscribed to the Google Groups "brms-users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to brms-users+unsubscribe@googlegroups.com.
To post to this group, send email to brms-...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/brms-users/4e47da1e-bb31-471a-ae53-f6ed74e3a43a%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

[Dimitris Rizopoulos]

Thanks for the clarification. Is there a way to obtain marginal predictions for hierarchical models with nonlinear link functions?

Στις Πέμ, 1 Μαρ 2018 στις 11:59 μμ ο χρήστης Paul Buerkner <paul.b...@gmail.com> έγραψε:

Yes indeed, only the fixed effects are used, which in non-linear models is not identical to marginal predictions. Unfortunately, the latter are far more difficult to compute...

In retrospect, the name "marginal_effects" was suboptimal. See also my comment in the details section of ?marginal_effects

2018-03-01 23:29 GMT+01:00 Dimitris Rizopoulos <dimitris....@gmail.com>:

I have a question for the predictions / fitted values produced by marginal_effects() for hierarchical models with random effects and nonlinear link functions (e.g., mixed effects logistic regression): Namely, in the help page of the function it is mentioned that when re_formula = NA (default), then marginal predictions are calculated. Does this mean that only the fixed-effects part of the model is used? For nonlinear link functions this wouldn’t correspond to marginal predictions, right?

Thanks for any pointers!

--
You received this message because you are subscribed to the Google Groups "brms-users" group.

To unsubscribe from this group and stop receiving emails from it, send an email to brms-users+...@googlegroups.com.

To post to this group, send email to brms-...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/brms-users/4e47da1e-bb31-471a-ae53-f6ed74e3a43a%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

--

Sent from Gmail Mobile

[Paul Buerkner]

There is no way built into brms, yet. You may open an issue on brms' github page, but I don't know when I will manage to implement it.

2018-03-02 7:15 GMT+01:00 Dimitris Rizopoulos <dimitris....@gmail.com>:

Thanks for the clarification. Is there a way to obtain marginal predictions for hierarchical models with nonlinear link functions?

Στις Πέμ, 1 Μαρ 2018 στις 11:59 μμ ο χρήστης Paul Buerkner <paul.b...@gmail.com> έγραψε:

Yes indeed, only the fixed effects are used, which in non-linear models is not identical to marginal predictions. Unfortunately, the latter are far more difficult to compute...

In retrospect, the name "marginal_effects" was suboptimal. See also my comment in the details section of ?marginal_effects

2018-03-01 23:29 GMT+01:00 Dimitris Rizopoulos <dimitris....@gmail.com>:

I have a question for the predictions / fitted values produced by marginal_effects() for hierarchical models with random effects and nonlinear link functions (e.g., mixed effects logistic regression): Namely, in the help page of the function it is mentioned that when re_formula = NA (default), then marginal predictions are calculated. Does this mean that only the fixed-effects part of the model is used? For nonlinear link functions this wouldn’t correspond to marginal predictions, right?

Thanks for any pointers!

--
You received this message because you are subscribed to the Google Groups "brms-users" group.

To unsubscribe from this group and stop receiving emails from it, send an email to brms-users+unsubscribe@googlegroups.com.

To post to this group, send email to brms-...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/brms-users/4e47da1e-bb31-471a-ae53-f6ed74e3a43a%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.