Sorry for the delay in replying, my back has been crook.
>Was this paper: RYLE, A. & EVANS, C.D.H (1991):
>SOME MEANINGS OF BODY AND SELF IN EATING-DISORDERED AND COMPARISON SUBJECTS
>BRITISH JOURNAL OF MEDICAL PSYCHOLOGY, 1991, Vol.64, No.Pt3, pp.273-283 ?
Yes
>Was the Bailey & Sims paper
>
>"THE REPERTORY GRID AS A MEASURE OF CHANGE AND PREDICTOR OF OUTCOME IN THE
>TREATMENT OF ALCOHOLISM"
>BRITISH JOURNAL OF MEDICAL PSYCHOLOGY, 1991, Vol.64, No.Pt3, pp.285-293 ?
Yes
>Question 2:
>When inteventions do seem very successful, do the effect sizes tend to be
>so big that there is little need to quibble about the statstical niceties?
>There seems at first sight to be all sorts of objections to treating
>distances in grids from different people with different consruct systems as
>if they provided a common metric. But if the effects are very big, maybe
>it/s not so bad. The Bailey & Sims idea recounted by Bob ( - haven't read
>their paper yet!) of standardising the distance betwen elements in some
>way, in terms of total variation in the grid, seems a good idea to me.
There are both philosophical and statistical issues involved in the issue of
measuring change via grids. Regarding your comments above, one problem will
be to agree on when an effect size is big enough, and the risk of ignoring
change because the effect wasn't perceived to be big enough. A further issue
concerns the importance of choosing an appropriate measure, as that the
effect size may well be influenced by the choice of the measure.
Regarding your comments:
>(You might have to use a general purpose stats package such as SPsS or
>Minitab for the PCA to do this - I'm not sure whether it's an option in the
>specialist grid packages). Then use Pythagoras's Theorem, in as many
>dimensions as there are factors, to calculate each distance. I haven't done
>such a calculation since high school, but it's really simple in principle
>and if I weren't in a tearing hurry I'd work it out here and now. Can any
>proper maths person chip in and confirm the answer?! Something like the
>square root of the sum of the squares of the differences between the scores
>on each factor ...
James Grice, referred me to Pythagora's theorem and sent me the following:
The formula is on page 91 of the 1957 edition (Osggod's Measurement of
Meaning). I don't think there is
a second edition. Essentially, for a pair of elements (E1, E2) and a
set of constructs (C1, C2, C3):
sqrt[(E1C1 - E2C1)**2 + (E1C2 - E2C2)**2 + (E1C3 - E2C3)**2]
sqrt = square root.
**2 = square
E1C1 = rating of element 1 on construct 1; E2C1 = rating of element
2 on construct 1; etc.
In words, sum the squared differences between the elements on
each construct. Then, take the square root of that sum. You can
compute this discrepancy score for any pair of elements, and
the distance between E1 and E2 will necessarily be the same as
the distance between E2 and E1 (i.e., order doesn't matter).
My interest at this stage is to stay as close to the data as possible hence
my interest in the issue of distance vs mean ratings vs sum of squares vs
correlations etc.
regards,
Bob
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%