Different results found at smartpls 2 and 3

This forum is closed, and read-only.
Locked
Jane
PLS Junior User
Posts: 4
Joined: Wed Oct 31, 2018 7:38 am
Real name and title: Ms. Jane Ling

Different results found at smartpls 2 and 3

Post by Jane »

Hi, i have different results found in my pls algorithm and bootstapping when i used both smart pls 2 and 3 especially on my mediating effect. The values for the independent variables are all matching just that for mediation to dependent variable and moderation to dependent variable values are very different. I have checked the missing value and etc but i couldn't figure out what is the cause. Wishing for some advice about this.

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

Re: Different results found at smartpls 2 and 3

Post by jmbecker »

Of what kind are the differences (paths, loadings, quality criteria, etc.)? Can you be more explicit?
Generally, there should be no differences in the basic estimates (path, loadings, weights) when you choose, the same (correct) settings. Also data treatment option should be the same (e.g., decimal delimiter).
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
Jane
PLS Junior User
Posts: 4
Joined: Wed Oct 31, 2018 7:38 am
Real name and title: Ms. Jane Ling

Re: Different results found at smartpls 2 and 3

Post by Jane »

I have recheck and found out the results are the same except still don't know why the path coefficient for moderating effect is different.
jmbecker
SmartPLS Developer
Posts: 1281
Joined: Tue Mar 28, 2006 11:09 am
Real name and title: Dr. Jan-Michael Becker

Re: Different results found at smartpls 2 and 3

Post by jmbecker »

First, please remind what is written on the website when downloading SmartPLS2: „SmartPLS 2.0.M3 was released in 2005. Since then there has been no maintenance or bugfix. Although it is very popular, SmartPLS 2.0.M3 contains some known issues that have never been fixed. You need to know these issues and some expert knowledge applying the necessary work arounds.“

The basic PLS and bootstrapping are fine, but as soon as you go to more advanced features there are a few known issues. You find some of them here: viewtopic.php?f=1&t=2113
One issue that is missing there, which is commonly discussed, deals with the moderating effects in SmartPLS 2. SmartPLS 2 standardizes all effects, but interaction effects need to be unstandardized to be correct (you will find many references to that problem in the literature). SmartPLS 3 correctly calculates the moderating effect leaving it unstandardized. Therefore, the estimates in SmartPLS 3 are correct without further ado, while the ones in SmartPLS 2 would need some extra transformation to be correct. So don't trust the results in SmartPLS 2 when it comes to moderating effects.
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
Jane
PLS Junior User
Posts: 4
Joined: Wed Oct 31, 2018 7:38 am
Real name and title: Ms. Jane Ling

Re: Different results found at smartpls 2 and 3

Post by Jane »

Sorry for the late reply. Noted with thanks.
Jane
PLS Junior User
Posts: 4
Joined: Wed Oct 31, 2018 7:38 am
Real name and title: Ms. Jane Ling

Re: Different results found at smartpls 2 and 3

Post by Jane »

May i double check the formula for f squared is it [(R square included - R square excluded)/(1- R square included)]? I tried to do manual calculation for f squared then found out the answer is different from the f squared i got in smart pls 3.
jmbecker
SmartPLS Developer
Posts: 1281
Joined: Tue Mar 28, 2006 11:09 am
Real name and title: Dr. Jan-Michael Becker

Re: Different results found at smartpls 2 and 3

Post by jmbecker »

Your formula seems correct. But note that what you calculate on your own is not what is calculated by SmartPLS. SmartPLS calculates the f² effect size based on a given model varying the regressors in the structural model regressions, but does not change the model by actually excluding constructs. The last would affect the whole model and thereby might produce strange outcomes for the f².
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
Locked