weights in 2nd oder constructs with repeated indicator appr.

Questions about the implementation and application of the PLS-SEM method, that are not related to the usage of the SmartPLS software.
Post Reply
badneben
PLS Junior User
Posts: 2
Joined: Mon Apr 14, 2008 8:40 pm
Real name and title:

weights in 2nd oder constructs with repeated indicator appr.

Post by badneben »

A thousand thanks for the software and this very helpful forum, however, having screened through the 2nd order LV discussions here

viewtopic.php?t=112
viewtopic.php?t=145
viewtopic.php?t=441

I haven't found an answer to the following burning question:

Huber/Herrmann/Meyer/Vogel/Vollhardt (2007) ("Kausalmodellierung mit Partial Least Squares"; Gabler/Wiesbaden) state that Wold's (1982. S. 40ff) hierarchical component approach is only valid, when the factors constituting the 2nd order construct have identical numbers of indicators.

Otherwise -they say- there will be a bias due to the no. of indicators of the 1st order LV that influences the weigths.

I have an all reflective modell with 4 LV (3 of them 2nd oder) that in turn have sometimes very different no. of indicators within their respective 1st level factor structure (2/7).

I tested both: (1) hierarchical components/repeated indicators and (2) using SPSS-calculated factors as manifest indicators of 2nd order LV.

The results are similar however differ.

Is there this above mentioned bias? Then I guess I'm better of using (2)?!
Post Reply