Applying weights to individual cases

This forum is closed, and read-only.
Locked
JohnP
PLS Junior User
Posts: 8
Joined: Sat Nov 14, 2009 9:48 pm
Real name and title:

Applying weights to individual cases

Post by JohnP »

There has been talk about the issue/feature before but I wish to find out if any progress has been made during the time passed since the last post about the subject.

When doing applied research and one wishes to weight the cases in the sample to reflect the distribution of the population - is this possible?

Former posts about the subject:

viewtopic.php?t=484&highlight=weight+data

viewtopic.php?t=212&highlight=weight+data

viewtopic.php?t=211

viewtopic.php?p=201&highlight=#201
John Pedersen

Cand.merc.emf

Copenhagen Business School
Vicki Juel Isaksen
PLS Junior User
Posts: 1
Joined: Mon Feb 13, 2012 5:33 pm
Real name and title:

Weigth data and still use SmartPLS

Post by Vicki Juel Isaksen »

I have the same question. I would like to weight my data and use SmartPLS for modelling.

Any tips how to do this?

Or do I need to model on non-weighted data?
jmbecker
SmartPLS Developer
Posts: 1282
Joined: Tue Mar 28, 2006 11:09 am
Real name and title: Dr. Jan-Michael Becker

Post by jmbecker »

If you want to weight each and every variable and not a specific construct or regression then simply do the following as a pre-processin step before importing the data into SmartPLS:

You probabaly have a weight for each individual i that you want to apply on each variables measured for this i, hence

newX(ij) = X(ij)*sqrt(weight(i)) for all variables Xj of each individual i.
JohnP
PLS Junior User
Posts: 8
Joined: Sat Nov 14, 2009 9:48 pm
Real name and title:

Post by JohnP »

Thank you Dr. Jan-Michael Becker,

But I'm not quite sure I can follow your method for weighting the cases. From what I can see and understand, your alteration to the data before import changes the values for each case, but it does not affect the weight of each case.

Ex:

Case Gender v1 v2 Weight v1_new v2_new
Case1 Female 4 3 2 5,66 4,24
Case2 Male 4 3 1 4 3

If females are underrepresented in the sample I would like to weigh the female cases higher. In ex double as much as males. With your method, is it correct to assign these new values to case 1 for the variable v1 and v2 (new ones)?

In my belief this is not functional but please enlighten me.

/John
John Pedersen

Cand.merc.emf

Copenhagen Business School
User avatar
Diogenes
PLS Super-Expert
Posts: 899
Joined: Sat Oct 15, 2005 5:13 pm
Real name and title:
Location: São Paulo - BRAZIL
Contact:

Post by Diogenes »

Hi,

This issue is very important, but is not implemented in SmartPLS.

About weighting the sample (not weighting variables), see http://essedunet.nsd.uib.no/cms/topics/weight/4/

They use two kind of weighting, and one of them is named
PWEIGHT
This weight corrects for the fact that most countries taking part in the ESS have very similar sample sizes, no matter how large their population. Without weighting, any figures combining two or more countries’ data would be incorrect, over-representing smaller countries at the expense of larger ones. The Population size weight makes an adjustment to ensure that each country is represented in proportion to its population size.

Best regards,

Bido
JohnP
PLS Junior User
Posts: 8
Joined: Sat Nov 14, 2009 9:48 pm
Real name and title:

Post by JohnP »

Thank you Professor Bido,

I'm sorry to hear that this is not implemented in SmartPLS, but that was what I expected to hear. Just wanted a clarification.

And it is not in any way possible to work around it by manipulating the data before import?

/John
John Pedersen

Cand.merc.emf

Copenhagen Business School
Locked