Rue Cromwell wrote:
----begin quote----
Dear Andy:
This is a fascinating research problem. Let me encourage you to pursue it
whether or not the following suggestions are helpful.
I presume, although I am not sure, that the matrix of responses of the child
client/subject are binary, i.e., 1 and 0, or yes and no, or check and void. If
not, e.g., if on a broader rating scale, they could be transformed to a binary
dimension by collapsing. Assuming, then, this binary matrix, I would suggest
that one additional binary variable be added which was not part of the child's
responses, i.e., whether this child did or did not have emotional problems, yes
or no.
Then, I would suggest applying de Boeck (Louvan, Belgium) & Rosenberg's
(Rutgers University) HICLAS analysis to the matrix. This yields an
hierarchical cluster analysis to each individual child's protocol. In addition
to the obvious interest in how these responses are grouped and hierarchically
arranged in the HICLAS analysis, it will also be of interest whether the
identification of the child as with or without emotional problems relates to
the hierarchical structure. My hunch is that the clusters which are
independent of the added variable will be of as much or more interest as the
variables which indeed bear upon the presence or absence of disturbance.
If you are not familiar with de Boeck's HICLAS program, I suspect you can get
the most recent version from Paul de Boeck (perhaps also from Seymour
Rosenberg). I do not have the addresses handy.
Best wishes for your good research progress.
Rue L. Cromwell
cromwell@kuhub.cc.ukans.edu
University of Kansas, where else?
----end of quote----
My response to the points you raise is:
Yes,the child's responses will produce a matrix with family members /
friends along one axis and indiviual items / questions from the test along
the other. So a typical example might be:
Family member / friend: M F S1 S2 SE F1 F2 where:
M = mother
Questions 1 X O X X 0 0 X F = father
from 2 X X X O O O O S = sibling
the 3 O 0 X O X X X SE = self
test 4 O X X O O X O F = friend
5 X O O X O O X
...... and so on .......
Question 1 might for instance be designed to elicit a response from the
child in which (s)he indicates which family members / friends (s)he
thinks love him a lot. Another question might ask who he or she gets
angry with alot, and so on.
I will need to spend along time thinking about the analysis of this.
Thank you for your comments on that and the correction that you mailed
me directly.
Christel Woodward wrote:
Hve you thought of comparing the results in a child relations test with
the family relations test? The child is allowed to nominate as many family
members as (s)he wishes to each construct, but the contructs used are
constrained by the test.
Chris Woodward, Ph.D.
Professor, Department of Clinical Epidemiology and Biostatistics
Faculty of Health Sciences
McMaster University, Hamilton, Ontario Canada L8N 3Z5
Phone: (905) 525-9140 ex.22131 FAX: (905)-546-5211
Yes, I know the Bene-Anthony Family Relations Test very well. Infact, it
was coming across this test in a clinical context that I first became
interested in this area. Unfortunately, this test has serious limitations.
It was first developed in 1957 and even in the most recent reprint (1985)
the language is often outmoded. I have recently conducted a semi-structured
telephone questionnaire on over 50 clinicians and researchers who have used,
or still use, the test. Many of them have similar misgivings and I hope to
publish the reslts of this as soon as I have finished the detailed analysis.
The items in the test rely too heavily on behavioural aspects of relationships
rather than how the child feels about a person. They also score items such
as "This person likes to tickle me" and being in bed with someone as strongly
positive, which may be true for some children, but certainly bot for those who
have been sexually abused.
Thank you again,
Andy
---------------------------------------------------------------------------
Andrew Parkin: Greenwood Institute of Child Health, Univ. of Leicester, UK.
email: ap23@leicester.ac.uk