Date: Tue, 22 Mar 2005 13:06:41 -0500
Sender: "SPSSX(r) Discussion" <SPSSX-L@LISTSERV.UGA.EDU>
Subject: Re: satisfaction study
Content-type: text/plain; charset=US-ASCII
Although Renji did not mention it specifically, multicolinearity of survey
variables used as independent predictors is often another complicating
issue when modeling overall satisfaction. PCA can often resolve this issue
while still providing meaninful insights for the business decision makers.
Market Research Manager
(804) 747-0422 x4831
(804) 935-4549 FAX
The information in this email is extremely confidential. It is intended
solely for the addressee. Access to this email by anyone else is
unauthorized. Please do not copy or disseminate any portion of this email.
Sent by: To
Re: satisfaction study
Please respond to
> Hello Renji, I'm not a statistician but maybe it makes sense to apply
> following process I saw long time ago:
> Try to apply Principal Component Factor Analysis with all predictors,
> scores for each individual and then try to explain overall
> based on those scores (seen as a combination of original variables)
> with OLS
> regression analysis.
Firstly principal component and Factor analysis are 2 differnt methods
under the parent 'Data reduction Mehods'.
You would only do PCA when you have huge number of independent variables
and don't know which one and how to specify in the model. factor scores
would give you the comprehensive bahavior of meaningful variables
loaded heavily on the principal component(s) extarcted.
In this case the problem is with the dependent variable which is
categorical in natute. you cannot do OLS regression. moreover even if
you combine the independent variables by PCA (as Victor said) you're
dependent variable is still categorical.
Logistic regression is the solution.