If you prefer I'll re-post this after Berlin.
When selecting the Euclidean distance/cluster analysis option in Gridstat,
the output includes a matrix of element (and construct) distances.
If I have two grids with the same elements and want to 'measure' if
construing of key elements has changed from grid 1 to grid 2 (I know this
involves a host of assumptions), is it statistically sound to use these
element distances to 'measure change'? In relation to Devi's post, you
referred to squared-euclidean-distance. Is this also applicable to the
above instance or does it exclusively relate to the factor example?
Other options for assessing change include:
mean ratings
cosine or correlation between element or construct
size of the sum of squares of an element compared to the average sum of squares
analyis of changes in ranks, rather than using ratings (an option Bill
Chambers posted to me).
In your view which is the generally preferred option?
regards,
Bob
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%