Converting a continuous moderator to categorical

Frequently asked questions about PLS path modeling.
Post Reply
LPhillips
PLS Junior User
Posts: 3
Joined: Fri Aug 05, 2016 9:37 am
Real name and title: Dr Laura Phillips

Converting a continuous moderator to categorical

Post by LPhillips » Thu Jun 06, 2019 6:48 am

Please could someone help me interpret this? – in my model, if I treat a moderator as continuous, it does not have a significant interaction effect, despite having a significant effect on the DV and its presence in the model reducing the effect of the IV on the DV and increasing the R squared. If I transform it, using cluster analysis, into a categorical variable (high, medium and low) and use PLS-MGA analysis it does have a significant moderating effect? Why would it not have an effect as a continuous variable - measurement invariance?

Secondly, could someone help explain to me the circumstances under which is it acceptable to convert a continuous moderator to a categorical variable? I have seen many argue against it.

Many thanks in advance.

jmbecker
SmartPLS Developer
Posts: 990
Joined: Tue Mar 28, 2006 11:09 am
Real name and title: Dr. Jan-Michael Becker

Re: Converting a continuous moderator to categorical

Post by jmbecker » Thu Jun 06, 2019 8:17 am

There is an interesting special issue in the Journal of Consumer Psychology (https://www.sciencedirect.com/science/a ... 0815000807) that deals with this issue. It is nothing particularly specific to PLS, but more a general question when it is ok and when not.
Dr. Jan-Michael Becker, University of Cologne, SmartPLS Developer
Researchgate: https://www.researchgate.net/profile/Ja ... v=hdr_xprf
GoogleScholar: http://scholar.google.de/citations?user ... AAAJ&hl=de

LPhillips
PLS Junior User
Posts: 3
Joined: Fri Aug 05, 2016 9:37 am
Real name and title: Dr Laura Phillips

Re: Converting a continuous moderator to categorical

Post by LPhillips » Fri Jun 07, 2019 9:46 am

Thank you for your response, this was very helpful.

After reading these articles, and reading about moderating effects in Henseler and Fassott's (2010) chapter in the Handbook of Partial Least Squares, I believe I am not seeing a significant interaction term (when using the product term approach) because the moderating effect is not linear. I believe I have a cascading moderating effect, where there are other categorical variables that are effecting the strength of the moderator effect. Theoretically this makes sense, my question is how do I model this in PLS-SEM? Do I use the product term approach for the direct effects and the product term for the main moderator effect and then run a PLS-MGA to see if the categorical variables moderate the interaction term? The Henseler and Fassott (2010) chapter only covers how to model cascading moderating effects where all variables are continuous.

Many thanks again in advance.

Post Reply