[Rasch] One model for three constructs

Juanito Talili talilij at yahoo.com
Mon Jun 6 15:19:42 EST 2011

Thank you for your reply, Mike.
Follow-up questions:
1.  You said, organize the items on the item map according to their strands.  What do you mean? Organize according to color coding (for instance, red, blue, etc)?
2.  I am worried because some items have large (>2.0) mean square fit.  According to Sick ( http://jalt.org/test/sic_5.htm), high mean square fit may be an indication that the item is measuring a different construct.  Does it mean that the item may measure a construct other than the three constructs which I analyzed together?
3. What is the command in order to save the rasch residuals?
4. I will do factor analysis using the residuals? Is it possible that the items will wildly loaded to unexpected constructs?  If so, what shall I do?
Thank you.

--- On Mon, 6/6/11, Mike Linacre <rmt at rasch.org> wrote:

From: Mike Linacre <rmt at rasch.org>
Subject: Re: [Rasch] One model for three constructs
To: rasch at acer.edu.au
Date: Monday, 6 June, 2011, 3:49 AM

Thank you for your questions, Juanito.

You asked: What shall I do with my three constructs? How to justify my boss that the three constructs are considered as one dimension?

Reply: Are your three constructs truly three dimensions (as different as height, weight, temperature)? Or are they three strands (as similar as addition, subtraction, multiplication in an arithmetic test)? If your three constructs all relate to the same thing in the same way, then they are probably three strands. If so, organize the items on the item map according to their strands.

You wrote:

Further, I followed your suggestion that I should do a Principal Component Analysis of the Residuals with Winsteps, then do a "shop" DIF analysis for the items by comparing the four shops in pairs.  My question is: What does it mean when DIF test is significant for several items  and nonsignificant for other items?  Can I still use the instrument for the four shops in the future ?
Reply: Let's think about an example. Suppose the manager of one shop emphasizes "friendliness". Then "friendliness" will be rated relatively highly by customers of that shop. So, in a DIF report, the "friendliness" item will have significant DIF for that shop. We would be unlikely to change the instrument or its administration because of this DIF. Instead, this DIF becomes part of our report about the shop. We want the same instrument to be administered in the same way in all 4 shops, so that we can see that "friendliness" performance is relatively higher (or the "friendliness" item is relatively easier) in one shop.

You wrote: Using Winsteps, how to construct customer-item map in four columns, one for each shop?

Reply: Here is one way. Do separate analyses of each of the four shops. Output the four person-item maps (Table 1). Then use the rectangular-copy function in Word (alt+mouse) to place the maps next to each other.


Mike L.

Mike Linacre
Editor, Rasch Measurement Transactions
rmt at rasch.org www.rasch.org/rmt/ Latest RMT:  24:4 Spring 2011 
-------------- next part --------------
An HTML attachment was scrubbed...
URL: https://mailinglist.acer.edu.au/pipermail/rasch/attachments/20110605/20afd782/attachment.html 

More information about the Rasch mailing list