[Rasch] multiple occasions

Mark Moulton markhmoulton at gmail.com
Sat Feb 13 07:34:20 EST 2010


I also like the "stacked approach".  It brings more information into the
calibration of the items, and thus a higher precision model.  Rasch
difficulty parameters are designed not to be test or administration

That said, I would validate the stacked approach by also doing separate
calibrations, then graphing the item difficulties against each other
(standardize first, so they have the same mean and SDs).  They should form
straight-line relationships.  If they don't, there may be a
test/administration interaction effect that needs to be checked out and may
require you to drop some items.  After you have cleaned out the "wandering"
items, I would do the final item calibration using "stacked" data.

Mark Moulton
Educational Data Systems

On Mon, Feb 8, 2010 at 2:02 PM, Petroski, Greg <
PetroskiG at health.missouri.edu> wrote:

>  I am still wrestling with the problem of scoring a single instrument
> administered on multiple measurement occasions to the same individuals.  It
> seems there are three options.
> 1.  The “stacked approach” in which data from K occasions from N
> individuals is stacked and items are calibrated and “ability” estimated as
> though there are NxK individuals.  This is easy to do and it would seem like
> the item parameters are then the average parameter values over all
> occasions.  So it is an appealing idea & easy to do.
> 2.  Estimate item parameters from the first occasion then estimate ability
> for all occasions as though the items were known.  Does this sound
> plausible?
> 3.  Separate calibrations, followed by an equating step.  This is
> technically the least pleasant option, but does it have some merit that I am
> missing?
> Greg Petroski
> U. of Missouri
>  -------------------------------------------------
> Please consider the environment before you print
> _______________________________________________
> Rasch mailing list
> Rasch at acer.edu.au
> https://mailinglist.acer.edu.au/mailman/listinfo/rasch

Please consider the environment before you print
-------------- next part --------------
An HTML attachment was scrubbed...
URL: https://mailinglist.acer.edu.au/pipermail/rasch/attachments/20100212/6a712116/attachment.html 

More information about the Rasch mailing list