[R-sig-ME] dropping interaction terms

[Sorkin, John]

Cristiano,


Your question has no absolutely correct answer.


If you have reason to believe that there is an interaction, even though your model does not demonstrate a significant interaction, you could justify leaving the interaction in the model and doing your post hoc tests. If you believe that the result you obtained regarding the interaction, that the interaction is not significant, then it would be most appropriate to drop the interaction, re-run the model and do your post-hoc tests.


Why would you want to do post hoc tests in a model that contains an interaction, an interaction that the model indicates is not significant, unless you have other data that lead you to believe that the interaction represents physical truth? If you have other data suggesting that the interaction does in fact represent reality (even though your data do not support this) you might want to include the interaction in you post hoc tests to as to most accurately model reality.


The above represents on old statistician's opinion. Like much in statistics the answer to your question reflects that fact that much of statistics is art rather than science.


John


John David Sorkin M.D., Ph.D.
Professor of Medicine
Chief, Biostatistics and Informatics
University of Maryland School of Medicine Division of Gerontology and Geriatric Medicine
Baltimore VA Medical Center
10 North Greene Street
GRECC (BT/18/GR)
Baltimore, MD 21201-1524
(Phone) 410-605-7119
(Fax) 410-605-7913 (Please call phone number above prior to faxing)



________________________________
From: R-sig-mixed-models <r-sig-mixed-models-bounces using r-project.org> on behalf of Cristiano Alessandro <cri.alessandro using gmail.com>
Sent: Wednesday, June 6, 2018 1:05 PM
To: Ben Bolker
Cc: R Mixed Models
Subject: Re: [R-sig-ME] dropping interaction terms

CAUTION: This message originated from a non UMB, UMSOM, FPI, or UMMS email system. Whether the sender is known or not known, hover over any links before clicking and use caution opening attachments.



It sounds like the question is controversial. Yes, I'll post it on cross
validated. Thanks!

On Wed, Jun 6, 2018 at 11:34 AM, Ben Bolker <bbolker using gmail.com> wrote:

>
>   This is really a broad linear modeling question rather than one that
> applies specifically to mixed models.  Can I suggest that you try
> CrossValidated [https://stats.stackexchange.com] ? (Be prepared for a
> variety of opinions ...)
>
>   cheers
>     Ben Bolker
>
> On 2018-06-06 12:45 PM, Cristiano Alessandro wrote:
> > Dear all,
> >
> > I have a simple design with two factors (each with two levels) and their
> > interaction term. I am analyzing it with mixed models.
> >
> > The interaction term is not significant. Do I need to fit another model
> > without the interaction term before performing post-hoc tests, or is it
> > more appropriate to perform post-hoc tests on the full model with the
> > non-significant interaction term?
> >
> > Thanks for your help
> >
> > Best
> > Cristiano
> >
> >       [[alternative HTML version deleted]]
> >
> > _______________________________________________
> > R-sig-mixed-models using r-project.org mailing list
> > https://stat.ethz.ch/mailman/listinfo/r-sig-mixed-models
> >
>
> _______________________________________________
> R-sig-mixed-models using r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/r-sig-mixed-models
>

        [[alternative HTML version deleted]]

_______________________________________________
R-sig-mixed-models using r-project.org mailing list
https://stat.ethz.ch/mailman/listinfo/r-sig-mixed-models

	[[alternative HTML version deleted]]