FIMIX PLS segments and MGA

Questions about the implementation and application of the PLS-SEM method, that are not related to the usage of the SmartPLS software.
Post Reply
huan
PLS Junior User
Posts: 2
Joined: Sat Aug 08, 2020 10:14 pm
Real name and title: Mr. Huan Bin

FIMIX PLS segments and MGA

Post by huan »

Dear everyone,

I have a model for identifying factors affecting environmental perception, and have divided the data (total number is 184) into two groups (group1 is 105 and group 2 is 79) based on different scenarios.
The MGA results shown that there are no significant differences between the path coefficients of the two groups, but through Bootstrapping, three path coefficients are significant in group 1, and one path coefficient is significant in group 2. It seems quite strange.
After read several articles, I realized maybe the unobserved heterogeneity effect existing in this study. And I try to run the FIMIX
The result of FIMIX with the total data shows that

one segment
AIC (Akaike's Information Criterion) 1455.5378
AIC3 (Modified AIC with Factor 3) 1466.5378
AIC4 (Modified AIC with Factor 4) 1477.5378
BIC (Bayesian Information Criteria) 1490.9021
CAIC (Consistent AIC) 1501.9021
HQ (Hannan Quinn Criterion) 1469.8714
MDL5 (Minimum Description Length with Factor 5) 1720.3593
LnL (LogLikelihood) -716.7689
EN (Entropy Statistic (Normed))
NFI (Non-Fuzzy Index)
NEC (Normalized Entropy Criterion)

two segments
AIC (Akaike's Information Criterion) 1430.1663
AIC3 (Modified AIC with Factor 3) 1453.1663
AIC4 (Modified AIC with Factor 4) 1476.1663
BIC (Bayesian Information Criteria) 1504.1098
CAIC (Consistent AIC) 1527.1098
HQ (Hannan Quinn Criterion) 1460.1366
MDL5 (Minimum Description Length with Factor 5) 1983.8839
LnL (LogLikelihood) -692.0832
EN (Entropy Statistic (Normed)) 0.7751
NFI (Non-Fuzzy Index) 0.8234
NEC (Normalized Entropy Criterion) 41.3877

Segment 1: 0.8851
Segment 2: 0.1149

Here are my questions:
1. Should I redivide the data into two analysis groups based on the results of FIMIX? But I don't know which data goes into group one and which goes into group two. And if I do that, how do I compare the data between different scenarios?
2. I have read in the link viewtopic.php?f=12&t=16122&p=26942&hilit=fimix#p26942 that if I'm already known my groups, maybe I don't need to run FIMIX. Therefore, how can I interpret this strange result, or identify potential heterogeneity?
jmbecker
SmartPLS Developer
Posts: 1282
Joined: Tue Mar 28, 2006 11:09 am
Real name and title: Dr. Jan-Michael Becker

Re: FIMIX PLS segments and MGA

Post by jmbecker »

1) Your sample size is quite small for dividing the data into groups. Especially, your second group is quite small. That could explain the non significant effects in that group and also why your MGA does not show significant differences.

2) The sample size limitation also applies to FIMIX. Especially as it shows a quite imbalanced distribution of observations to groups. Your second group from FIMIX would certainly have not enough observations to reliably estimate a MGA.

3) To discover whether there is a threat with unobserved heterogeneity you need to estimate a sequence of models from only one group to several groups and compare the statistics.

The following references could be of interest do you:
Becker, J.-M., A. Rai, C. M. Ringle, and F. Völckner (2013). "Discovering Unobserved Heterogeneity in Structural Equation Models to Avert Validity Threats." MIS Quarterly, 37(3), 665-694.

Hair, Joe F., M. Sarstedt, L. Matthews, and C. M. Ringle (2016). Identifying and Treating Unobserved Heterogeneity with FIMIX-PLS: Part I – Method, European Business Review, 28 (1), 63-76.

Matthews, L., M. Sarstedt, J. F. Hair, and C. M. Ringe (2016). Identifying and Treating Unobserved Heterogeneity with FIMIX-PLS: Part II – A Case Study, European Business Review, 28 (2), 208-224.

Sarstedt, M.; Christian M. Ringle, and Hair, J. F. (2018). Treating Unobserved Heterogeneity in PLS-SEM: A Multi-method Approach. In H. Latan & R. Noonan (Eds.), Partial least squares structural equation modeling: Basic concepts, methodological issues and applications. New York: Springer.
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
Post Reply