R help - Jun 2009 - Conducting data modelling on weighted data using R

Hello,

I am starting to use R for various analyses, for example I use the ca package to
do Correspondence Analysis.  I am also looking to use packages such as:

pls  	Partial Least Squares
plspm    Partial Least Squares  Path Modelling

However, although I can use packages such as these on un-weighted data there
does not appear to be a facility to take account of weighted data.

I am a statistician working for a Market Research company where in many projects
I work on weighted data. Typically, the data will be structured in the following
manner:

Case_id	y	x1	x2	x3	x4 	weight
1		5	3	4	1	4	0.345
2		3	4	3	1	5	0.543
3		2	2	3	3	3	1.456
4		4	4	2	4	3	0.345
5		3	3	4	3	2	1.200
.
.
etc

The above data is fictional.    Where y is the dependent variable, and there are
4 independent variables (x1, x2, x3 and x4),  and weight is the weighting factor
variable.

For some functions in R such as lm   (Linear Regression) the functionality is
provided to conduct linear regression on weighted data:

e.g.  model <- lm(y ~ x1+x2+x3+x4, weights=weight)


I was just wondering if anyone had any suggestions of how to apply packages such
as pls and plspm on weighted data?  Or should I contact the Maintainers of the
packages to see if they can include a weights command?

Many Thanks for any assistance in this area.

Kind Regards

Peter





Peter Harding | Senior Methodologist | Marketing Sciences & Advanced
Analytics | Harris Interactive
O: +44 (0)20 8263 5249 |  M: +44 (0)7890 561391 | E:
pharding@harrisinteractive.com
Watermans Park, High Street, Brentford, TW8 0BB, UK
www.harrisinteractive.com/europe

Remember that we have a global online omnibus service at your disposal. 25+
markets run as frequently as weekly, and priced from £195 per question. 
Flexibility of 1,000 or 2,000 people each time.  See our omnibus in action in
the FT, Herald Tribune, the Metro and many more.







	[[alternative HTML version deleted]]