Hi,
I have a question regarding PLS and factor analysis
To check the dimensionality of my latent constructs, I have performed a factor analysis in SPSS. The factor analysis in SPSS found 8 different factors, which is conceptually correct. Furthermore, the factors were checked by Cronbach's Alpha.
When I use these 8 factors in SmartPLS, one of the factors (latent constructs) do not have significant factor loadings from none its indicators.
Could someone explain to me, why this is the case?
To my understanding the PLS technique is partly a factor analysis, so one could expect to have the same dimensions in PLS as in a SPSS factor analysis…
Best regards
Michael
Factor analysis and PLS
-
Diogenes
- PLS Super-Expert
- Posts: 899
- Joined: Sat Oct 15, 2005 5:13 pm
- Real name and title:
- Location: São Paulo - BRAZIL
- Contact:
Hi Michael,
do you have used extraction with common factor or principal components? sometimes the first is better for this porpose.
rotation was oblique? sometimes the last factors show this kind of problem because, in fact, they have lower loadings than the first ones.
Best regards.
Bido
do you have used extraction with common factor or principal components? sometimes the first is better for this porpose.
rotation was oblique? sometimes the last factors show this kind of problem because, in fact, they have lower loadings than the first ones.
Best regards.
Bido
Last edited by Diogenes on Mon Jan 07, 2008 7:58 pm, edited 1 time in total.
Hi Bido,
I have used Principal Component as extraction method and Varimax rotation with Kaiser Normalization. I have tried other extraction methods, which gives a somehow different constellation of the latent constructs (with only 6 constructs instead of the conceptually 8).
I am not sure if I understand, what you mean by: “Sometimes the last factors show this kind of problem because, in fact, they have lower loadings than the first ones”…
Most of all, I am looking for a theoretically explanation to why the difference in the results?
Many thanks
-Michael
I have used Principal Component as extraction method and Varimax rotation with Kaiser Normalization. I have tried other extraction methods, which gives a somehow different constellation of the latent constructs (with only 6 constructs instead of the conceptually 8).
I am not sure if I understand, what you mean by: “Sometimes the last factors show this kind of problem because, in fact, they have lower loadings than the first ones”…
Most of all, I am looking for a theoretically explanation to why the difference in the results?
Many thanks
-Michael
-
Diogenes
- PLS Super-Expert
- Posts: 899
- Joined: Sat Oct 15, 2005 5:13 pm
- Real name and title:
- Location: São Paulo - BRAZIL
- Contact:
Hi Michael,
In EFA, usually, even after rotation, the first factors have more indicators and bigger loadings than the last factors. Remembering that all factors have loading in all indicators.
When we translate this factors as measurement model in PLS, usually we connect the factors with the indicators that had bigger loading in EFA, but now, the LV will be estimated with only these indicators. If the last factors had lower loadings could happen the problem that you are reporting.
Just thinking..
Best regards
Bido
In EFA, usually, even after rotation, the first factors have more indicators and bigger loadings than the last factors. Remembering that all factors have loading in all indicators.
When we translate this factors as measurement model in PLS, usually we connect the factors with the indicators that had bigger loading in EFA, but now, the LV will be estimated with only these indicators. If the last factors had lower loadings could happen the problem that you are reporting.
Just thinking..
Best regards
Bido
-
viswadatta
- PLS Expert User
- Posts: 139
- Joined: Wed Jul 05, 2006 1:43 pm
- Real name and title:
- Location: Coimbatore, India
pls vs factor analysis
PLS is CFA, while factor analysis usually refers to EFA. In factor analysis the end result gives the components of each factor. In PLS we assume it apriori and also check for construct validity.