Path models with manifest variables PLS?

Frequently asked questions about PLS path modeling.
Post Reply
winkler

Path models with manifest variables PLS?

Post by winkler »

Dear all,

are there any experiences in path modeling without latent variables (structures directly between manifest variables)? Pros and Cons or other methods for this specific kind of models?

Looking forward to any remarks or ideas on this question.

Best Regards

Nadine
winkler

Post by winkler »

Hi Philipp,

thanks for your hint. If I understood correctly the post refers to the subject of including objective measures (or binary scaled variables). The manifest variables I want to analyse are scaled (5 point) items, so subjective measures.

My question is linked to the last part of the post: What about measuaring structures between the items directly not including latent variables (circles).
cgrimpe
PLS User
Posts: 11
Joined: Tue Nov 15, 2005 12:20 pm
Real name and title:

Post by cgrimpe »

Hi Nadine,

Not necessarily binary variables. "Objective" just stands for anything that refers to a "hard" measure like turnover, number of patents etc. So no problem with 5-point-scales.

But anyway, I would also be highly interested in your initial question whether a structural model made up of single manifest variables instead of latent variables would do.

Best
Christoph
User avatar
Diogenes
PLS Super-Expert
Posts: 905
Joined: Sat Oct 15, 2005 5:13 pm
Real name and title:
Location: São Paulo - BRAZIL
Contact:

Re: Path models with manifest variables PLS?

Post by Diogenes »

Hi,

with SEM covariance based you can look at
BOLLEN, Kenneth A. Structural Equations with latent variables. John Wiley & Sons, 1989. (cap. 4 = SEM with observed variables).

with SEM PLS based you can look at
http://www.hec.unige.ch/recherches_publ ... 002.25.pdf (they compared PLS regression and OLS regression).

Using Smartpls is easy, you just have to put each variable as a formative indicator of one LV, i. e.,

each LV will have just one formative indicator.

Best regards.
Prof. Dr. Diogenes de Souza Bido
stefanbehrens
PLS Expert User
Posts: 54
Joined: Wed Oct 19, 2005 5:53 pm
Real name and title:

Post by stefanbehrens »

Hi all,

I agree. The only way to do this in SmartPLS is to create an LV (bubble) for each indicator (square). Since the loading and weight will always be 1, it doesn't even matter whether you choose a reflective or formative measurement mode.

However, if all your variables are measured by a single indicator, why even use PLS in the first place? Multiple regression in any statistics suite (e.g., SPSS) would do just fine...

Regards,
Stefan
User avatar
Diogenes
PLS Super-Expert
Posts: 905
Joined: Sat Oct 15, 2005 5:13 pm
Real name and title:
Location: São Paulo - BRAZIL
Contact:

Re: Path models with manifest variables PLS?

Post by Diogenes »

Hi, just to inform...

MOROCO, João. Análise estatística com utilização do SPSS. Potugal, Edições Sílabo, 2003.

in pages 457-476 the path analysis using SPSS (only MV) is explained.

The model is calculated with multiple regression (one equation to each endogenous variable)

Best regards.
Prof. Dr. Diogenes de Souza Bido
Post Reply