[Rasch] Rasch analysis: repeated measures
liasonas at lycos.com
Mon Feb 13 03:57:27 EST 2006
I have the Language school assessment results of around 7000 pupils. The school year is split in three terms. Therefore, for each child I have the "Term 1 grade", "Term 2 grade", "Term 3 grade" (sale 1-20, only integers). Each child also takes a test at the end of the year, so I have a "Test score" (which is a fourth score - imagine this to be a fourth column in a repeated measures design - again on a 1-20 scale). The first three measures of ability are given by the teacher (but are usually based on informal classroom tests). Therefore, the fourth one is slightly different in nature in the sense that it is a formal large-scale external test. We may assume that this is a situation of 'repeated measures design' with four measures that measure change in Language ability. How can I use Rasch analysis to analyse these results?
Option a: Run four different indepentent analysis, then get the ability measures and run a Repeated Measures ANOVA
Option b: Run a 3Faceted Rasch analysis where each of the four measures is a different Rater. This assumes that the progress/learning/change effects will be absorbed by the leniency of the raters, right?
Option c: to perform a single analysis by treating the children measured on different occasions as distinct ones (so the sample size will be 4 X N). Then use the four measures for a repeated measures ANOVA
Could you please indicate which of the above methods (or another new one that is not included above) is more appropriate? Are there any publications / methodological papers on how to use Rasch analysis in this context? I have the impression that there is a lack of knowledge. Has anyone read a paper comparing the above methods?
Any contribution will be appreciated
> ----- Original Message -----
> From: "Mike Linacre (RMT)" <rmt at rasch.org>
> To: rasch at acer.edu.au
> Subject: [Rasch] More - multiple comparisons - BH method
> Date: Tue, 07 Feb 2006 14:03:26 -0600
> Thanks everyone for your further input.
> Benjamini has published numerous papers on the BH method and
> modifications (some dated 2006). It seems he still recommends the
> original BH method for general-purpose use, even though it does
> have a slight conservative bias (much smaller bias than Bonferroni
> - too conservative, or stand-alone - too liberal).
> For DIF studies, a slight conservative bias is probably beneficial.
> Precision of computation is no problem with modern computers.
> Mike Linacre
> Rasch mailing list
> Rasch at acer.edu.au
Dr. Iasonas Lamprianou
CFAS, School of Education
The University of Manchester
Oxford Road, Manchester M13 9PL, UK
Tel. 0044 161 275 3485
iasonas.lamprianou at man.ac.uk
Search for businesses by name, location, or phone number. -Lycos Yellow Pages
More information about the Rasch