Dear SmartPLS users/experts/developers,
Can anyone please explain to me how to calculate the q-squared (NOT Q-squared)? I am not sure how to run the blindfolding procedure to get the Q-squared excluded because I have no idea which LV shall I select/check. I would appreciate if you could explain this issue to me with an example.
Thank you very much.
Preditive relevance q-squared
-
Hengkov
- PLS Super-Expert
- Posts: 1599
- Joined: Sun Apr 24, 2011 10:13 am
- Real name and title: Hengky Latan
- Location: AMQ, Indonesia
- Contact:
Hi Ida,
Please see two chapter books below for detail calculate and example: effect size (f-squares), Q-squares and q-squeres predictive relevance.
Chin, W.W (1998). The partial least squares approach to structural equation modeling. In G.A. Marcoulides (Ed.), Modern methods for business research (pp. 295-336). Mahwah, New Jersey: Lewrence Erlbaum Associates.
Chin, W.W (2010). How to write up and report PLS Analyses. In V.E Vinsi et al (Eds.), Hanbook of partial least squares: Concepts, methods and applications in marketing and related fields (pp. 655-690). Berlin: Spinger.
Regards,
Hengky
Please see two chapter books below for detail calculate and example: effect size (f-squares), Q-squares and q-squeres predictive relevance.
Chin, W.W (1998). The partial least squares approach to structural equation modeling. In G.A. Marcoulides (Ed.), Modern methods for business research (pp. 295-336). Mahwah, New Jersey: Lewrence Erlbaum Associates.
Chin, W.W (2010). How to write up and report PLS Analyses. In V.E Vinsi et al (Eds.), Hanbook of partial least squares: Concepts, methods and applications in marketing and related fields (pp. 655-690). Berlin: Spinger.
Regards,
Hengky
How do you use blindfolding for Q² correctly?
If you select all latent variables in your model via Blindfolding menu, will it calculate the right CV red. for each construct or do you have to select each single construct to calculate its individual CV red?
Moreover, regarding the predicive "size" (q²), I understand, that the method to measure that is the same as for effect size. Do you have to temporarily delete one predictor to get the influence on a subsequent construct?
Any suggestions would be really appreciated!
Best,
Julius
If you select all latent variables in your model via Blindfolding menu, will it calculate the right CV red. for each construct or do you have to select each single construct to calculate its individual CV red?
Moreover, regarding the predicive "size" (q²), I understand, that the method to measure that is the same as for effect size. Do you have to temporarily delete one predictor to get the influence on a subsequent construct?
Any suggestions would be really appreciated!
Best,
Julius
Hi Hengky,
Thanks for the advice on the literature. However, I was a little more interested in the technique within SmartPLS.
My first problem is, when I run the blindfolding over all construcst and I look at a specific construct, the value differs from the one I get when I just select this construct in the menu.
Which one is the true Q² score? The single-measured CV red. or the CV red. including all constructs?
As for q², the way I did that was analogous to the effect size measurement. But this time, the blindfolding procedure has a menue where you can actually choose constructs. Shall I delete each predictor and measure the CV red afterwards and compare these scores?
Best,
Julius
Thanks for the advice on the literature. However, I was a little more interested in the technique within SmartPLS.
My first problem is, when I run the blindfolding over all construcst and I look at a specific construct, the value differs from the one I get when I just select this construct in the menu.
Which one is the true Q² score? The single-measured CV red. or the CV red. including all constructs?
As for q², the way I did that was analogous to the effect size measurement. But this time, the blindfolding procedure has a menue where you can actually choose constructs. Shall I delete each predictor and measure the CV red afterwards and compare these scores?
Best,
Julius
Last edited by Joules on Thu Jun 28, 2012 1:16 pm, edited 2 times in total.
Could anyone of you confirm if that answers my questions above?
I found a statement by Prof. Ringle here viewtopic.php?t=944&highlight=blindfolding
He says, to measure the Q² for an individual endogenous variable, one must select only the construct of interest in the blindfolding menu and unselect all other constructs.
If that's right, I have to repeat that procedure for all other endogenous variables.
For q², I then temporarily delete each predictor variable step by step and run the same blindfolding procedure all over. For each predictor the change in Q² of the endogeous variable is then q²=Q²incl-Q²excl/ (1- Q²incl). (same method as for R² changes)?
Is that correct?
I'd be grateful for any advice!
Julius
I found a statement by Prof. Ringle here viewtopic.php?t=944&highlight=blindfolding
He says, to measure the Q² for an individual endogenous variable, one must select only the construct of interest in the blindfolding menu and unselect all other constructs.
If that's right, I have to repeat that procedure for all other endogenous variables.
For q², I then temporarily delete each predictor variable step by step and run the same blindfolding procedure all over. For each predictor the change in Q² of the endogeous variable is then q²=Q²incl-Q²excl/ (1- Q²incl). (same method as for R² changes)?
Is that correct?
I'd be grateful for any advice!
Julius
-
christian.nitzl
- PLS Expert User
- Posts: 248
- Joined: Sat Jul 25, 2009 1:34 pm
- Real name and title:
-
spiking
- PLS Expert User
- Posts: 47
- Joined: Wed Aug 05, 2015 3:35 am
- Real name and title: Clemen Chiang
Re: Preditive relevance q-squared
Is there any online calculator to determine q square?
Thank you.
Thank you.
-
Hengkov
- PLS Super-Expert
- Posts: 1599
- Joined: Sun Apr 24, 2011 10:13 am
- Real name and title: Hengky Latan
- Location: AMQ, Indonesia
- Contact:
Re: Preditive relevance q-squared
Hi,
Q-square value will always be almost equal to the value Adjsuted R ^ 2. If you have a model that is simple, with only one dependent variable, then the value of both will be identical.
Regards,
Q-square value will always be almost equal to the value Adjsuted R ^ 2. If you have a model that is simple, with only one dependent variable, then the value of both will be identical.
Regards,
-
spiking
- PLS Expert User
- Posts: 47
- Joined: Wed Aug 05, 2015 3:35 am
- Real name and title: Clemen Chiang
Re: Preditive relevance q-squared
It will be helpful if there is an Excel template that we can use to generate small q square.Hengkov wrote:Hi,
Q-square value will always be almost equal to the value Adjsuted R ^ 2. If you have a model that is simple, with only one dependent variable, then the value of both will be identical.
Or how about generating small q square from SmartPLS?
Any available?