[R] Significance test, Cronbach's Alpha

[Tom Backer Johnsen]

falissard wrote:
> This test can be meaningfull, but only in rare circumstancies (are the items
> (or raters) more consistent than what could be expected by chance?).

Do you have a reference to such a test?

> I would do it with R using a bootstrap procedure (Efron & Tibshirani, An
> introduction to the bootstrap, p.156).

I agree.  That should be possible, actually quite simple.  Run through 
the data a large number of times, randomize the sequence of the 
observations for each case and compute alpha for each run.

Tom
+----------------------------------------------------------------+
| Tom Backer Johnsen, Psychometrics Unit,  Faculty of Psychology |
| University of Bergen, Christies gt. 12, N-5015 Bergen,  NORWAY |
| Tel : +47-5558-9185                        Fax : +47-5558-9879 |
| Email : backer at psych.uib.no    URL : http://www.galton.uib.no/ |
+----------------------------------------------------------------+