Many people interested in these kinds of questions, and who have
developed methods for dealing with them are members of these two lists.
Society for Mathematical Psychology: MPSYCH Listserv
<http://aris.ss.uci.edu/smp/mpsych.html>
Classification Society of North America CSNA
<http://www.pitt.edu/%7Ecsna/index.html>
class-l link
Art
[log in to unmask]
Social Research Consultants
University Park, MD USA
(301) 864-5570
Rosie McEachan wrote:
>Dear Allstat,
>
>I am currently in the process of designing a questionnaire based study
>and desperately need some advice about what type of statistics to use.
>The questionnaire will ask a number of individuals (i), to rate a number
>of dimensions (j), on a number of different behaviours (k). The aim of
>the study is to discover a) which behaviours are seen as more similar
>(i.e. which group together), and b) which dimensions differentiate
>between these behaviours (so that we could say behaviour 1 is high on
>dimension 1, but low on dimension 2 etc). Since I am asking about
>perceptions of behaviours there will be individual variation in
>responses, however, I am more interested in getting a general picture of
>how behaviours relate to one another than I am about investigating
>individual differences. Not being a statistician myself, I am unsure of
>the best way to statistically analyse this type of 3 mode data, so was
>after some advice. So far I have uncovered two options.
>
>1) I could use cluster analysis to cluster a matrix of behaviour x
>dimensions, where each entry would have to be a summary measure of all
>the individuals in the sample. According to Everitt, Landau & Leese
>(2004, Cluster Analysis) one way to do this would be to use
>Mahalanobis's distance as a summary measure, as this would take into
>account some of the variability between individuals (more so than a mean
>score, for example).
>2) I could use a form of three-way component analysis (e.g. Kiers
>& Mechelen, 2001, Psychological Methods, 6, 84-110), perhaps 3MPCA
>(three mode factor analysis), which as far as I understand could come up
>with factor solutions for all three modes, and represent their
>interactions
>
>I was hoping that some of you might be able to shed light of which one
>of these would be more appropriate to my study. I would welcome any
>suggestions, or references for further reading (especially introductory
>chapters) in this area. Please reply to me at [log in to unmask]
>Many thanks,
>Rosie McEachan
>
>Rosie McEachan
>Research Student
>Institute of Psychological Sciences
>University of Leeds
>Tel: 0113 343 6692
>(internal ext 36692)
>Fax: 0113 343 5749
>
>
>
>
>
|