The effect of correlation on the formation of clusters can ... - GreenBook
The effect of correlation on the formation of clusters can ... - GreenBook
The effect of correlation on the formation of clusters can ... - GreenBook
Create successful ePaper yourself
Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.
Collinearity could be a problem in <strong>the</strong> regressi<strong>on</strong> models, but it’s<br />
not clear how much <str<strong>on</strong>g>of</str<strong>on</strong>g> a problem it would be in <strong>the</strong> formati<strong>on</strong> <str<strong>on</strong>g>of</str<strong>on</strong>g><br />
<strong>the</strong> segments. SOM is a neural network that makes virtually no<br />
assumpti<strong>on</strong>s about <strong>the</strong> data. While its reliance <strong>on</strong> Euclidean<br />
distances for creating a topological map could make it vulnerable<br />
to collinearity problems, o<strong>the</strong>r available distance measures may<br />
alleviate <strong>the</strong> problem. Methods such as CHAID rely <strong>on</strong> splitting<br />
variables <strong>on</strong>e at a time and hence may not be susceptible to<br />
collinearity problems. However, without a thorough investigati<strong>on</strong>,<br />
it’s impossible to say all <strong>the</strong>se methods are free <str<strong>on</strong>g>of</str<strong>on</strong>g> collinearity<br />
problems, even though it does appear that, in <strong>the</strong>ory, <strong>the</strong> problem<br />
should be less severe than in cluster analysis.<br />
Segmentati<strong>on</strong> by itself is a l<strong>on</strong>g and arduous process<br />
requiring choices to be made based <strong>on</strong> study objectives, type <str<strong>on</strong>g>of</str<strong>on</strong>g><br />
data, analytical method, and guidelines for decisi<strong>on</strong> criteria. It<br />
doesn’t need to be fur<strong>the</strong>r complicated by <strong>the</strong> presence <str<strong>on</strong>g>of</str<strong>on</strong>g><br />
collinearity in <strong>the</strong> data when cluster analysis is used to create <strong>the</strong><br />
segments. While it is true that practical segmentati<strong>on</strong> studies<br />
almost always use more than two variables, I hope <strong>the</strong> informati<strong>on</strong><br />
from this study will lead to a better understanding <str<strong>on</strong>g>of</str<strong>on</strong>g> <strong>the</strong> problems<br />
caused by collinearity and possible soluti<strong>on</strong>s. Of course, this<br />
preliminary effort needs to be expanded in scope to properly<br />
understand <strong>the</strong> role <str<strong>on</strong>g>of</str<strong>on</strong>g> collinearity in segmentati<strong>on</strong> problems.<br />
About <strong>the</strong> Author<br />
Rajan Sambandam is vice president <str<strong>on</strong>g>of</str<strong>on</strong>g> research at <str<strong>on</strong>g>The</str<strong>on</strong>g> Resp<strong>on</strong>se<br />
Center in Fort Washingt<strong>on</strong>, Penn. He may be reached at<br />
rsambandam@resp<strong>on</strong>se-center.com.<br />
Reprinted with permissi<strong>on</strong> from <strong>the</strong> Ameri<strong>can</strong> Marketing Associati<strong>on</strong> (Marketing Research, Vol 15, No. 1, Spring 2003)<br />
21