I have a significant 3-way multilevel model interaction with the lmer package in r. When I probe this interaction using the online tool from Preacher, then interaction has regions of significance. In contrast, if I use the interactions package, simple_slopes function, there are no regions of significance. Anyone have an idea why there would be differences and which one might be correct?
4
-
$\begingroup$ Good on you for trying both, at first I thought the interactions package wouldn't support mixed-effects models but from the documentation it looks like it does. Also it looks like they both use the methods outlined in Bauer & Curran (2005). So I too am excited to see what could explain this difference. One difficulty with the Preacher tool is entering every parameter estimate without error--did you double and triple check your entries? (Both in terms of their exact values and which parameters you're entering) $\endgroup$– Patrick CoulombeCommented Dec 7, 2020 at 18:11
-
$\begingroup$ @PatrickCoulombe. Thanks for the response. Yes, I've entered the values twice and got the same result. I might try again just to be sure. Even if I entered the data incorrectly, do you know what could account for a significant 3-way interaction and then null regions of significance if the interactions package output is correct? $\endgroup$– BenCommented Dec 7, 2020 at 18:14
-
$\begingroup$ no I don't know, but wait, are you saying that there's no regions of significance provided or that the regions are provided but there is no values for which the simple slopes are significant?! that doesn't seem possible... If the former, then I think it's got to do with lmer not providing confidence intervals, which is referred to in the documentation of the interactions package $\endgroup$– Patrick CoulombeCommented Dec 7, 2020 at 21:42
-
$\begingroup$ It's providing conditional effects at different levels of the moderator, but these are not significant. In addition, the plot shows that there are no significant regions. $\endgroup$– BenCommented Dec 9, 2020 at 14:14
Add a comment
|