Defining interaction in random effects in lme4

Previous Topic Next Topic
 
classic Classic list List threaded Threaded
2 messages Options
Reply | Threaded
Open this post in threaded view
|

Defining interaction in random effects in lme4

Dominik Ćepulić
Dear everybody!

My fixed-effects-only model looks like this: glmer(Accuracy ~ C.RT*Group,
data = da)

C.RT is the reaction time variable, and Group is a categorical variable
with 0 and 1 as values. I would like to specify that main intercept, Group
intercept, C.RT slope and C.RT*Group slope vary across subjects and trials.

All subjects have values in Group = 0 and in Group = 1. Trials are nested
within Group because each trial belongs either to Group = 0 or Group = 1.
How should I specify the model?

My ideas were:

   1. glmer(Accuracy ~ C.RT*Group + (C.RT*Group|subject) + (1+C.RT|trial),
   data = da)

or
    2. glmer(Accuracy ~ C.RT*Group + (1+C.RT|Group:subject) +
(1+C.RT|Group:trial), data = da)

Here, Group:trial does not make much sense as trials are *per se* divided
in Group 0 or Group 1.

What is, in your opinion, the best way to specify the model that I want to
test?

Additionally, the difference between (1+C.RT|Group:subject) and
(C.RT*Object|subject) is not clear to me. Can someone also shed some light
here?

Thanks,
Dominik!

        [[alternative HTML version deleted]]

______________________________________________
[hidden email] mailing list -- To UNSUBSCRIBE and more, see
https://stat.ethz.ch/mailman/listinfo/r-help
PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
and provide commented, minimal, self-contained, reproducible code.
Reply | Threaded
Open this post in threaded view
|

Re: Defining interaction in random effects in lme4

Bert Gunter-2
Probably better to post this on the r-sig-mixed-models list.

Cheers,
Bert



On Jan 7, 2018 12:20 PM, "Dominik Ćepulić" <[hidden email]> wrote:

> Dear everybody!
>
> My fixed-effects-only model looks like this: glmer(Accuracy ~ C.RT*Group,
> data = da)
>
> C.RT is the reaction time variable, and Group is a categorical variable
> with 0 and 1 as values. I would like to specify that main intercept, Group
> intercept, C.RT slope and C.RT*Group slope vary across subjects and trials.
>
> All subjects have values in Group = 0 and in Group = 1. Trials are nested
> within Group because each trial belongs either to Group = 0 or Group = 1.
> How should I specify the model?
>
> My ideas were:
>
>    1. glmer(Accuracy ~ C.RT*Group + (C.RT*Group|subject) + (1+C.RT|trial),
>    data = da)
>
> or
>     2. glmer(Accuracy ~ C.RT*Group + (1+C.RT|Group:subject) +
> (1+C.RT|Group:trial), data = da)
>
> Here, Group:trial does not make much sense as trials are *per se* divided
> in Group 0 or Group 1.
>
> What is, in your opinion, the best way to specify the model that I want to
> test?
>
> Additionally, the difference between (1+C.RT|Group:subject) and
> (C.RT*Object|subject) is not clear to me. Can someone also shed some light
> here?
>
> Thanks,
> Dominik!
>
>         [[alternative HTML version deleted]]
>
> ______________________________________________
> [hidden email] mailing list -- To UNSUBSCRIBE and more, see
> https://stat.ethz.ch/mailman/listinfo/r-help
> PLEASE do read the posting guide http://www.R-project.org/
> posting-guide.html
> and provide commented, minimal, self-contained, reproducible code.
>

        [[alternative HTML version deleted]]

______________________________________________
[hidden email] mailing list -- To UNSUBSCRIBE and more, see
https://stat.ethz.ch/mailman/listinfo/r-help
PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
and provide commented, minimal, self-contained, reproducible code.