🚀 go-pugleaf

In <emu-ct08.cg.1993.0702.123416.nacd@tuda.ncl.ac.uk> Anthony Downing <A.C.Downing@NEWCASTLE.AC.UK> writes:

>In a simulation, one has a
>God's-eye-view, so I knew these dichotomous items were tapping the single
>underlying continuum a little above or below its mean, but what if the
>data had come from real life?  Look at the frequency splits of most
>published questionnaires with dichotomous items, and you will find them
>far more skewed than the items in my simulation, yet find factor analysis
>usually being invoked to say what they are measuring!

Too true.   If you go on to run simulations using a variety of cut
points and varying amounts of introduced error, but always making sure
that you have three or more `items' representing any cut point, you
will discover just what a fickle mistress factor analysis can be.
Skewness in more-or-less continuous distributions can be just as
troublesome.   I did all this back in 1977 or 1978, and it was the
experience of being able to defeat any of the available criteria for
number of factors that led me to doubt the wisdom of placing any
credence in item factoring at all.   But just try explaining any of
this thread to an enthusiast for whom f.a. holds out the prospect of
magical insight.

Anyone else remember the X-Ray spectacles advertised in Superman
comics in the 1950's?   Curious how they worked through dresses but
not underwear.   But I digress.   What possible connection could there
be with factor analysis...... ;-)

-- 
Steve Blinkhorn <steve@prd.co.uk>