|
|
Hello all,
On the back of a market research survey I have created a market
segmentation using a k-means clustering technique. If I then have
additional respondents that I wish to allocate to appropriate segments,
I consider I have two possible ways of doing this:-
1. Use a discriminant analysis, or
2. (assuming we have responses to the full set of questions used to
create the segmentation in the first place) assign each respondent to a
cluster according to the shortest distance to each cluster centre.
Under what circumstances would the former approach be more relevant than
the latter? Are there any problems or difficulties with using the
second method?
Many thanks in advance,
Mark
|
Locked
