>
.... deletion (see other response!)
>
> Regarding your interest in discussion of grid analysis I will
> offer a few comments as I took a somewhat different and critical
> approach to this subject. In research I undertook, SPSS was used
> to perform multidimensional scaling on 40 grids. To interpret the
> output 1)graphs of various constructs were used to determine if
> the dimensional interpretations which I made separated the
> elements in a similar manner, 2)SPSS principal component analyses
> of individual grid data were used to see if individuals with the
> largest weights on a particular dimension had similar types of
> constructs which loaded on the principal component. Also output
> from G-Pack, i.e., construct correlations, was used to check
> interconnections between constructs. For example, a person might
> have constructs which did not seem relevant to the interpretation
> which was made, though the constructs might all be highly
> correlated to another construct which was the link between the
> constructs.
>
> I believe the above approach enhanced the validity of the
> dimensional interpretations ... though apologise to people who
> are not familiar with the "fun" of mds as my brief description
> of method has been limited. In formulating my use of mds I found
> some of the feedback I received from contributors to this list
> most helpful, particularly as there is very little in the grid
> literature on mds and related approaches.
>
> Regards,
> Bob Green
>
I thought I'd put up a very quick response as there seems to be some
traffic developing about grids (goody!) MDS is like PCA in that it
extracts lower dimensional approximations to the grid data. Nicer in
that it can be applied to multiple grids fairly easily (using ALSCAL
in SPSS or SAS) and in that the data can be treated as ordinal (with
or without allowing untying of ties) rather than implicitly as ratio.
Only slight drawbacks are that it's computationally intensive and
therefore difficult to implement in specific grid packages. Bob's
work sounds very thorough and impressive illustrating other ways of
looking into the meanings of common dimensions extracted by MDS by
mathematical means (at least, I think that's what Bob was doing!)
Chris
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%