Q-square predictive relevance test

Frequently asked questions about PLS path modeling.
Post Reply
mkuppusa
PLS Junior User
Posts: 5
Joined: Mon Feb 16, 2009 11:09 pm
Real name and title:

Q-square predictive relevance test

Post by mkuppusa » Thu Jun 03, 2010 6:36 am

Dear PLS users,

Would anyone have some idea how to interpret Q-square results from SmartPLS blindfolding test? Do we look at the 1-SSE/SSO output or something else? Appreciate some tips/guidance.

Thanks
Mudiarasan Kuppusamy

christian.nitzl
PLS Expert User
Posts: 248
Joined: Sat Jul 25, 2009 1:34 pm
Real name and title:

Post by christian.nitzl » Thu Jun 03, 2010 3:43 pm

Dear Kuppusamy,
with Q-square you are testing the prediction relevance of your model. Q-square values above zero indicated that your values are well reconstructed and that the model has predictive relevance. You find the right output in SmartPLS at the “Construct Crossvalidated Redundancy” Output. There you have to look at the block with the “Total “ sums. However, as reported elsewhere in this forum, the blindfolding procedure has a bug at the actual release (s. viewtopic.php?t=944). You have pay attention to it. Further theoretical explanations for Q-square you can find in: Henseler, J./Ringle, C. M./Sinkovics, R. (2009): The Use of Partial Least Squares Path Modeling, in: Advances in International Marketing, Vol. 20 on pages 303 to 305.

I hope this helps!

Best regards

Christian

iris_afandiphd
PLS Senior User
Posts: 28
Joined: Thu Jun 24, 2010 5:38 am
Real name and title:

Post by iris_afandiphd » Wed Feb 23, 2011 1:22 am

Dear Christian,

Thank you for a great explaination
My questions are:

1. How many omission distance needs for running BF
2. I got a big number on TOTAL for every cases, it really more than 0

"Cross-validated R-square (i.e., Stone-Geisser’s Q2) between each endogenous latent variable and its own manifest variables can be calculated automatically in SmartPLS, Stone-Geisser’s Q2 by blindfolding and R2 by running the PLS procedure (Chatelin et al. 2002), and more than 0 is substantial..

Thanks

christian.nitzl
PLS Expert User
Posts: 248
Joined: Sat Jul 25, 2009 1:34 pm
Real name and title:

Post by christian.nitzl » Wed Feb 23, 2011 9:18 am

Please check following post:

viewtopic.php?t=1532&highlight=integer

There you can find some information about the omission distance.

Greetings,

Christian

iris_afandiphd
PLS Senior User
Posts: 28
Joined: Thu Jun 24, 2010 5:38 am
Real name and title:

Post by iris_afandiphd » Thu Feb 24, 2011 3:18 pm

Hi Christian,

Thank you for replying.
Do you mean this table

Image

Thank you

christian.nitzl
PLS Expert User
Posts: 248
Joined: Sat Jul 25, 2009 1:34 pm
Real name and title:

Post by christian.nitzl » Thu Feb 24, 2011 4:40 pm

Hey,

this isn`t the right table. You have to check “construct cross-validated redundancy” instead of “indicator cross-validated redundancy”.

Best regards,

Christian

ruchi
PLS Expert User
Posts: 75
Joined: Wed Dec 22, 2010 3:09 pm
Real name and title:

Post by ruchi » Thu Feb 23, 2012 4:03 am

1 )what does cases mean here? I have data of 85 respondents and one of my Latent variable has 3 indicators (manifest variables). Then in this, what will be my omission distance.

2) What about Formative Latent Variables ( whos indicators are formative). Can I calculate Q^2 for this formative construct also.

3) what about my LV with only one indicator

samaro
PLS Junior User
Posts: 7
Joined: Sat Mar 20, 2010 10:32 am
Real name and title:

Post by samaro » Thu Feb 07, 2013 3:36 pm

I think it has to do with the omission distance...
Suzanne Amaro

samaro
PLS Junior User
Posts: 7
Joined: Sat Mar 20, 2010 10:32 am
Real name and title:

Post by samaro » Thu Feb 07, 2013 3:37 pm

Henseler et al. (2009) said that blindfolding procedures is only applied to latent variables that have a reflective measurement model operationalization.
Suzanne Amaro

Post Reply