[Rasch] Differences in alpha for Winsteps and R

Stuart Luppescu lupp at uchicago.edu
Thu Jul 13 03:08:55 AEST 2017


On Wed, 2017-07-12 at 07:40 +1000, Mike Linacre wrote:
> The answer is "(approximate due to missing data)" !!
> 
> Cronbach Alpha is defined for complete data, so when data are
> missing, 
> assumptions must be made about those data. Winsteps computes Alpha
> as 
> though the data are complete, and ignores the fact there are missing 
> data, except for adding the "approximate" message.

Ah, I see. So, Mike, how does Winsteps compute Alpha "as though the
data are complete"? I believe that psych::alpha() does listwise
deletion. Sometimes if there is too much missing data it fails
completely.

> Stuart, what happens when you analyze a single form in Winsteps? Do
> the Alphas of Winsteps and psych::alpha() agree?

I haven't done that yet. I analyzed all forms together and then did a
post-hoc analysis of the Rasch measures by form. I'll try it and see
what we get. 

-- 
Stuart Luppescu
Chief Psychometrician (ret.)
UChicago Consortium on School Research
http://consortium.uchicago.edu

lupp at uchicago.edu


More information about the Rasch mailing list