🚀 go-pugleaf

This thread is a disgrace to stat-l.  Where has all the rigour gone?
We clearly need a proper experimental design here if we are all to
regain our dignity.   I suggest a repeated measures design (large
measures what is more, and of good quality single malt at that).   The
dependent variable would be the number of factors claimed to be
identified in a factor analysis of random data.   A learned discussion
on the effects of rotation after six pints of Old Peculiar could then
lead to the conclusion that further research was necessary.   Or as an
old colleague once put it, emphasising his insistence on setting all
non-significant correlations to zero before factoring a matrix:

"We all of us use Guttman, but some of us look at the stars!"
-- 
Steve Blinkhorn <steve@prd.co.uk>