Confusing differences in results using PLS-SEM vs LISREL

Frequently asked questions about PLS path modeling.
Post Reply
Andreas
PLS Junior User
Posts: 2
Joined: Fri May 25, 2018 8:26 am
Real name and title: Andreas H - PhD candidate

Confusing differences in results using PLS-SEM vs LISREL

Post by Andreas » Fri May 25, 2018 9:37 am

Hi all,

I have a dataset of N=263 of Likert scale questionnaire data that I got help from statistician to analyze using LISREL. I provided the raw data and the statistician cleaned the data and imputed some data to manage some of the missing data. When this was done the data set had 263 data points. The statistician then analyzed the data using LISREL (Maximum Likelihood) and my research model. Now, when I use the very same data and research model and analyse it using SmartPLS, some of the results are confusingly different. E.g. the coefficient between two certain LVs is high (-0.38) and significant (T-value -2.6) in the LISREL case. In contrast, the results from SmartPLS says that the very same coefficient is low (-0.04) and not significant (T-value 0.51).

Unfortunately, I do not have access to the statistician any more, but I do have the full analysis and I can repeat it myself using LISREL ad I get the same results. In the LISREL analysis, a few error terms are set to correlate with other item's errors. Also, one item for each LV is set as a reference level (item_x = 1* LV) etc. In the SmartPLS analysis, I do not do anything resembling that, but otherwise the conditions for the two analyses are the same from what I understand.

I am very confused that LISREL says that a relation is fairly strong and significant, while SmartPLS says that the same relation is very low and not significant. Now, what is it that I am missing here? Any pointers or suggestions are very welcome!

jmbecker
SmartPLS Developer
Posts: 820
Joined: Tue Mar 28, 2006 11:09 am
Real name and title: Dr. Jan-Michael Becker

Re: Confusing differences in results using PLS-SEM vs LISREL

Post by jmbecker » Fri May 25, 2018 1:26 pm

PLS and LISREL are somewhat similar, but not equivalent. You cannot expect results to be the same. Under some situations they can even strongly differ. And even within the CB-SEM (LISREL) models the results can vary strongly depending on the estimator (ML, GLS, ULS, etc.)

Our resources provide some first information about the differences between CB-SEM (LISREL) and PLS:
https://www.smartpls.com/documentation/ ... with-cbsem
https://www.smartpls.com/documentation/ ... and-cb-sem
https://www.smartpls.com/documentation/ ... d-when-not

The missing value treatment can also be a potential problem.
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

Andreas
PLS Junior User
Posts: 2
Joined: Fri May 25, 2018 8:26 am
Real name and title: Andreas H - PhD candidate

Re: Confusing differences in results using PLS-SEM vs LISREL

Post by Andreas » Sun May 27, 2018 8:14 pm

Thank you very much for the references! I will take a dive in them and hopefully get a lot wiser :) Thank you again!

Post Reply