Tim O'Connor

CSU (csu@brain.wph.uq.oz.au)
15 Feb 1996 13:20:22 -0500


I tried to send this to you personally but it came back to me:
I'm not a fan of cluster analysis but if you are looking at the
use of a grid for therapy purposes, you may find cluster analysis
useful- especially if you present the tree diagram back to the
person and ask them to confirm any interpretation or the
meaningfulness of results.

I can't answer your technical/statistical questions but the
validity of the analysis may lie in whether the results mean
anything to the person who completed the grid.

It might also be useful to plot the elements as columns as in a
graph, pick 1 construct of interest and make a graph (the ratings
would be the vertical axis/the elements the horizontal). If
useful you could do this for each construct... just another way
of presenting the results.

There is also the issue of the spatial plot from the PCA. Has
the person who completed the grid commented on this? Personally
I have found multidimensional scaling an interesting approach,
though it is not easy to grasp at first.

If you were interested I could put your data into G-Pack and see
what comes of it. To do this it would be useful to have the
construct poles. It would take me a few days to fit it in but
if its not urgent I'd be happy to give it a go,

Also check out: Shaw M (1980). The Analysis of a Repertory
Grid. British Journal of Medical Psychology. 53, 117-126.
I think a grid with data similar to yours was analysed.


Bob Green