Negative q2

This forum is closed, and read-only.
Locked
Renata3
PLS Junior User
Posts: 1
Joined: Mon Oct 31, 2016 3:29 pm
Real name and title: Renata

Negative q2

Post by Renata3 »

Dear all,
I have searched the web already but could not find an answer to my question.

My Q2 included and Q2 excluded are both positive, but since the value of Q2included is smaller, the q2 is negative.
I wondered what does it mean and how should I interpret it.

Thank you for your help.
jmbecker
SmartPLS Developer
Posts: 1281
Joined: Tue Mar 28, 2006 11:09 am
Real name and title: Dr. Jan-Michael Becker

Re: Negative q2

Post by jmbecker »

It could mean that your model is overfitting to the data and thus it has a better predictive releavance without the focal predictor.

However, there are quite many issues with the q² (small q). You need to consider that your context of the model changes and also that the data imputation in the blindfolding procedure might change.
This is a very under-researched field and thats why we do not provide an automatic calculation of the q² effect size.

I would also test if you find a similar result using the new PLS predict (comparing the RMSE and MAPE of the model with and without the predictor). If you find a similar result, then in fact, your model has better predictive power without the focal variable.
Dr. Jan-Michael Becker, BI Norwegian Business School, SmartPLS Developer
Researchgate: https://www.researchgate.net/profile/Jan_Michael_Becker
GoogleScholar: http://scholar.google.de/citations?user ... AAAJ&hl=de
lindaruit
PLS Junior User
Posts: 1
Joined: Fri May 05, 2017 10:01 am
Real name and title: Ms Linda Ruit

Re: Negative q2

Post by lindaruit »

Does a negative Q2 also mean that you cannot/do not have to calculate effect size q2?

Thanks!
User avatar
cringle
SmartPLS Developer
Posts: 818
Joined: Tue Sep 20, 2005 9:13 am
Real name and title: Prof. Dr. Christian M. Ringle
Location: Hamburg (Germany)
Contact:

Re: Negative q2

Post by cringle »

A negative Q² means that your model does not have predictive relevance. Why would you, then, compute the q² effect size? You can get the results but would it be of any use? Probably not too much.

Best regards
CR
Locked