[IPAC-List] Unidimensionality

[Doverspike,Dennis]

Good question but a complex and difficult one. One of the best discussions of the troublesome nature of this topic can be found in Lumsden, who gives your adding digits example. Of course, that is a very narrowly banded test.

There are both content-oriented and statistically-oriented definitions of unidimensionality, I would argue. 

I think your practically oriented is a good point. There are three dominant views of measurement - 1) fundmanetalist; 2) operationalist 3) funcitonal, practical, latent trait based. For most of us in psychology, in assessment, and in psychometrics, we tend to accept the practical, latent trait, view of measurement. So we tend to be practically oriented in our approach. 

This leads me to sort of my own five simple rules:
1. Do you believe that the underlying latent construct can be regarded as having a common core and thus being unidimensional?
2. Are you using a test construction approach consistent with unidimensionality? Some measurement approaches, for example according to Nunnally criterion-related test construction, are not consistent with a claim of unidimensionality. Of course, the debate would be whether much of the job knowledge testing we do is really unidimensional or multidimensional, but lets say we can make an adequate argument for unidimensional there.
3. Are you calculating a single total or overall score? If one score - unidimensional - if more than one score - multidimensional (or as with the Big Five composed of multiple unidimensional tests).
4. Is the validity and reliability data for the test consistent with unidimensionality? If i calculate two scores from the test instead of one do I get better reliability and validity?
5. If you calculate other statistics are they consistent with unidimensionality? 

I think unidimensionality is a lot like randomness? Unless we come right out and say - nope this is not unidimensional, we can take a stab at regarding it as unidimensional.       

Dennis Doverspike

________________________________________
From: ipac-list-bounces at ipacweb.org [ipac-list-bounces at ipacweb.org] On Behalf Of Shekerjian, Rene [Rene.Shekerjian at cs.state.ny.us]
Sent: Thursday, December 02, 2010 9:14 AM
To: ipac-list at ipacweb.org
Subject: [IPAC-List] Unidimensionality

I was wondering what folk's (your) understanding of unidimensionality was. I am guessing that most would agree that something like the ability to add single digit numbers without carrying would be pretty close to unidimensional. But past something like that, at what point does an ability become multidimensional?



Adding single digit numbers with carrying?

Dividing double digit numbers?

Calculating the standard deviation?

Driving a car?

Supervising staff?



And do you consider there to be both theoretical unidimensionality and practical unidimensionality? Or is there only one kind?



Thank you in advance,



René





René Shekerjian | Testing Services Division | NYS Department of Civil Service | 518-474-3778

=====================================================================



_______________________________________________________
IPAC-List
IPAC-List at ipacweb.org
http://www.ipacweb.org/mailman/listinfo/ipac-list