Dear All,
I have two queries related to factor analysis. Any comments would be
most appreciated.
1) I am trying to develop a statistical argument about why two
related constructs (risk-taking and self-harm) should be assessed
using the same instrument.
I thought that the use of a single measure consisting of items
sharing the same response format enables the investigation of the
factor structure of several RT and SH behaviors irrespective of
differences in scaling and distribution properties that could have
otherwise influenced the quality of the factor solution. Does this
make sense statistically and could anybody suggest a relevant
reference to back up this argument?
2) I have conducted an EFA and then a CFA and randomly split the
sample into two parts to avoid problems of capitalising on chance
fluctuations in the sample. Does this make statistical sense
and again is there a relevant reference I could use?
Many thanks
Ioanna
Ioanna Vrouva, PhD
Department of Clinical Psychology
University College London
1-19 Torrington Place
London WC1E 7HB
Email: [log in to unmask]
|