Article ID | Journal | Published Year | Pages | File Type |
---|---|---|---|---|
10525448 | Journal of Statistical Planning and Inference | 2005 | 10 Pages |
Abstract
Correspondence analysis appears to be sensitive to modalities with low margins leading to outliers. We use the notions of influence function and Hadamard differentiability to provide a criterion for deciding when such an outlier can be considered as too influential.
Related Topics
Physical Sciences and Engineering
Mathematics
Applied Mathematics
Authors
Emmanuel Nowak, Avner Bar-Hen,