Eins,
A couple of points:
1. The standard dichotomous Rasch model is nothing more than a simple
logistic regression, parameterized without a grand intercept. In fact,
one could actually fit a Rasch model in SPSS quite easily.
2. Providing guidelines with respect to "reporting scale development
and validation" would take quite a long time.
Here's a non-exhaustive list in no particular order off the top of my
head of what I tend to examine upon fitting a Rasch model:
Assuming you are dealing with standard dichotomous Rasch model, it
would be important to assess for:
1. item fit (infit and outfit) via mean-squares
2. person fit (infit and outfit) via mean-squares
3. item reliability & separation
4. person reliability & separation
5. dimensionality via an unrotated PCA on the probability scale
residuals, or perhaps a bifactor model
6. person-item map to examine person-item targeting, gaps,
distribution of persons and items, bandwidth, etc.
7. item discriminations by fitting a 2-PL model
8. global fit (e.g., test if there is a significant improvement
between the 2-PL and Rasch)
9. item characteristic curves (ICCs)
10. Differential item function via examination of ICCs
11. item information curves
12. test information curve
13. atypical response patterns (e.g., Depression: non-endorsement of
"mild" depressive items, yet endorsement of suicidal ideation)
...and the list goes on and on...
Several of the points listed above are inter-related. This list won't
mean much to those who are unfamiliar with Rasch modeling, but I think
this is a reasonable answer given the question and the forum.
Ryan
On Thu, Dec 6, 2012 at 10:51 PM, E. Bernardo <
[hidden email]> wrote:
> Dear Folks,
>
> In scale development and validation, some researchers use EFA and
> subsequently CFA (using different sample). This approach has clear
> guidelines for reporting Scale Development and Validation Results in the
> Journal of the Society for Social Work and Research (available at
>
http://www.jsswr.org/article/view/6974/4882 ).
>
> Sometime ago, there were exchanges in this list regarding the above
> procedure (EFA then CFA). Some folks do not favor the use of EFA. If you
> know references/guidelines for reporting scale development and validation
> without the EFA, please post them here.
>
> If you know also some guidelines for reporting scale development and
> validation using Rasch Model, please post them also here.
>
> Thank you.
> Eins
=====================
To manage your subscription to SPSSX-L, send a message to
[hidden email] (not to SPSSX-L), with no body text except the
command. To leave the list, send the command
SIGNOFF SPSSX-L
For a list of commands to manage subscriptions, send the command
INFO REFCARD
Locked
