Re: Help!!!

Richard Bell (
Mon, 20 Oct 1997 09:41:48 +1000

Dear Kathleen,

There are many different solutions to your problem, some of which have been
suggested by others. My only caution about Devi's solution however, is that
in taking construct as the unit of analysis you will need to be careful
(vis-a-vis your supervisor) of within-subject effects, since your
constructs will be 'nested' under respondents.

Continuing with a quantitative approach depends a little on the kinds of
grids you have collected. (I think I have a conference paper somewhere on
the analysis of multiple grids, a paper given many times to the same yawns
of enthusiasm, that I could send you if interested.) If the grids have
something in common, either elements or constructs (or both), then you
would be better off abandoning Flexigrid for a standard statistical
package, which can take advantage of the commonality (I have a large paper
on analysing grids through SPSS if you have access to that.)

If your grids are all different however, then you will need to stick with
Flexigrid, but use it as a pre-processor to generate summary data that can
subsequently analysed to produce the statistics beloved of 'research'
theses. There is not much written about this process though I have been
meaning to write a paper about for some time. Maybe it will be my essay in
boredom for Chester.

Hope this is of some help.

Richard Bell