Urgent - Multigroup Analysis

Frequently asked questions about PLS path modeling.

Urgent - Multigroup Analysis

Postby rainnl » Wed Apr 19, 2017 4:46 pm

Hello all,

I am currently in the process of analysing a combined-data set and in comparison both a female group data (group 1) set as well as a male group data set (group 2). While I seem to get significant relationships (t-value >1.96 in a 5%) between variables for the combined data, for the same relationship I keep getting non-significant values for both the female and male group - in both cases the t-value is hence lower than 1.96. So my question is simply: Is such an outcome possible and can occur? Please clarify and assist why this may happen with my data.

My common sense for some reason tells me that at least one of the two groups t-value should be greater than 1.96 and thus significant, given that the outcome for the combined data is significant. I am looking forward to receive assistance as quickly as possible, since I do not have much time left for this analysis. Thank you very much in advance!

Best regards,
Nicolas
rainnl
PLS Junior User
 
Posts: 4
Joined: Mon Apr 10, 2017 1:35 pm
Real name and title: Nicolas Luebbers

Re: Urgent - Multigroup Analysis

Postby rainnl » Thu Apr 20, 2017 6:05 pm

Anyone? Help / clarification would be appreciated. Thanks in advance!

Nicolas
rainnl
PLS Junior User
 
Posts: 4
Joined: Mon Apr 10, 2017 1:35 pm
Real name and title: Nicolas Luebbers

Re: Urgent - Multigroup Analysis

Postby jmbecker » Fri Apr 21, 2017 3:22 pm

You are right. Theoretically, the overall effect is a weighted average of group effects.

Thus, the most obvious reason would be lack of statistical power in both groups.
When you split your overall sample into subgroups you reduce the sample size (having roughly only half as much observations in both groups as on the overall sample). This reduction in sample size could lead to insufficient statistical power to detect the corresponding effect size. This would especially be a problem if the effect is only barely significant on the overall sample (i.e., a small effect).

As mentioned above: the overall effect is only theoretically the average of the group effects. Other potential problems could involve different multicollinearity patterns in the subgroups and overall sample that lead to suppression effects.
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
jmbecker
SmartPLS Developer
 
Posts: 692
Joined: Tue Mar 28, 2006 11:09 am
Real name and title: Dr. Jan-Michael Becker


Return to FAQ (Methodology)