<tentative interpretation deleted>
First, I will assume that you are interested in grouping the elements,
and so have used a correlation matrix or matching-score matrix based
on them, rather than on the constructs. Second, the problem with this
analytic method is that 6 of the constructs are elicited, meaning that
you have a majority of variables where nobody is measuring the
elements on the same scale. This will attenuate your similarity
matrix enormously by throwing lots of random noise into it. If you
eliminate the idiosyncratic constructs and use the four that were
supplied, I imagine that you'll get quite clean solutions, probably in
2 dimensions.
> the 3DMS is that there is greater separation of 5 elelements
> which are clustered together in the 2DMS. The first two
> dimensions in both options are highly correlated (.98 and .87,
...probably due to the randomness of the idiosyncratic constructs.
> and both solutions can be interpreted, though the third dimension
> has a less definite interpretation eg there are two possible
> interpretations. My inclination is to accept the two dimensional
> solution.
My own is to re-run it on a set of similarity matrices based on
*common* constructs, and my personal *hunch* is that your third
dimension is the one you really want, but it will only appear clearly
after cleaning up the data.
> Re 2) A regression using the 4 supplied constructs in the grid
> as dependent variables and the subject weights as independent
> variables did not result in any significant findings.
Again, because of the larger proportion of random variance due to
idosyncratic constructs being equated in your calculations.
> G-Pack and
> SPSS were used to examine high factor/pca loadings in individual
> grids which might suggest recurring important constructs. The
> other strategy has been to seek independent interpretation from
> experienced clinicans to confirm my interpretations.
MDSing individual grids is a *good* way to explore idiosyncracies in
the data, and is what I would recommend you do before interpreting the
INDSCAL on common constructs.
Best of luck!
((((((((((((((((((((((((((((((((((((((((((((((((((((((((((
Travis Gee () tgee@alfred.carleton.ca ()
() () ()()()()
() () ()
() ()()()()()()()()()
"In science, the more we know the more extensive the
contact with nescience." -Spencer