🚀 go-pugleaf

I have a faculty member who is doing a relability and validity study.
She is interested in performing an item analysis on data she has
collected.  She has several scales, each with about 103 items.  She
would like to reduce each scale to about 25 items.  Her N is about
175 (although not all participants in her survey answered all
questions).  After the items have been reduced she needs to validate
her new scales, then wants to see how here scales are correlated with
each other.

My first thought was to use a factor analysis to reduce the number of
variables, but the inter-item correlations are not especially large
(I visualize an "ill-conditioned matrix" warning on the horizon) and
the N is way to small to factor 103 variables.

My problem is that I have heard the term item analysis many times,
but I am totally unfamiliar with the analysis techniques used (this
wasn't covered in any of my stat courses). Does any one have any
suggestions?  I'll be happy to summarize discussion results for the
list.

Thanks,

Julie Scott
Research Computing Support		*** When all else fails   ***
University Computing Services		*** follow the directions ***
Western Michigan University