[Rasch] Rasch to G-theory

Mike Linacre rmt at rasch.org
Tue Apr 10 16:43:51 EST 2012


Jason:

The variances needed to compute a Rasch-based equivalent to a 
Generalizability Coefficient (G) are in Facets Tables 5 and 7.

My apologies that the "explained variance" computations at 
www.winsteps.com/facetman/table5.htm do not show the last step in the 
computation. It is:

G = (Explained variance% by target facet) / 100

Please verify this value of G makes sense for your data. A more specific 
Generalizability Coefficient may be more meaningful to your audience. 
Variance terms similar to 
http://www.stanford.edu/dept/SUSE/SEAL/Reports_Papers/ReliabCoefsGTheoryHdbk.pdf 
page 20 can be extracted from Facets output. These can be used to compute a 
more focused Generalizability Coefficient.

Mike L.


At 4/10/2012, you wrote:
>can anyone direct me to a paper which explains how to use Rasch measured 
>reported by Facets to compare the results with G-theory estimates? For 
>example, I can read in the Facets output the sd of item measures, the sd 
>of rater measures and the sd of person measures. How can I produce an 
>index which is similar to the variance attributed to markers as computed 
>in generalizability theory? Or maybe this is not possible? A good search 
>on the internet did not help me a lot, although I reached a paper by 
>Macmillan (2000) which, unfortunately, I cannot have access.

Mike Linacre
Editor, Rasch Measurement Transactions
rmt at rasch.org www.rasch.org/rmt/ Latest RMT: 25:4 Spring 2012
-------------- next part --------------
An HTML attachment was scrubbed...
URL: https://mailinglist.acer.edu.au/pipermail/rasch/attachments/20120410/00e29740/attachment.html 


More information about the Rasch mailing list