[Rasch] One model for three constructs and four groups (was RE: PCA and DIF)

Stuart Luppescu slu at ccsr.uchicago.edu
Wed Jun 15 04:23:13 EST 2011


On Sat, 2011-06-11 at 05:25 -0700, Juanito Talili wrote:
> Dear All,
> Stuart, you wrote: This is a tricky question. You have to balance the
> amount of mis-measurement with your desire to compare measurements
> across shops. REPLY: what is meant by "balance the amount of
> mis-measurement".? 

If you force item parameters to certain values that are different from
what they would be if you estimated them without restrictions, there
will be some misfit. Existence of misfit implies that you are not
measuring the people as well as you can. What I mean by "balance" is
that you have to accept some mis-measurement if you want to maintain
comparability across groups. If you want to measure the people as well
as possible, then you are going to lose some of the comparability. You
have to decide which is more important to you. Balance mis-fit with
comparability.
-- 
Stuart Luppescu -=- slu .at. ccsr.uchicago.edu        
University of Chicago -=- CCSR 
才文と智奈美の父 -=-    Kernel 2.6.38-gentoo-r6                
I strongly suggest you collaborate with a local
 statistician. I can think of no circumstance where
 multiple regression on "hundreds of thousands of
 variables" is anything more than a fancy random
 number generator.    -- Berton Gunter       R-help
 (February 2008)  
 
 
 
-------------- next part --------------
A non-text attachment was scrubbed...
Name: smime.p7s
Type: application/x-pkcs7-signature
Size: 3573 bytes
Desc: not available
Url : https://mailinglist.acer.edu.au/pipermail/rasch/attachments/20110614/424704fd/attachment.bin 


More information about the Rasch mailing list