Modified atkinson method: forward search algorithm
AuthorHamed, F. M. O.
MetadataShow full item record
CitationHamed, F. M. O. (2009). Modified atkinson method: forward search algorithm. Maltepe Üniversitesi. s. 158.
The method discussed in this paper is Atkinson’s Forward Search Algorithm (Atkinson, 1994). This is a powerful robust statistical method for detecting multiple outliers. Multiple outliers can have a strong influence on the model fitted to data. The Forward Search seeks to distinguish a larger ”clean” part, which is called the ”good” class, from outlier data, the ”bad” class. When there are two separated groups in the data, it rejects one of the groups in favour of the other. The main strategy is to separate ”good” from ”bad” data, where the ”good” data lie in one of the clusters and the ”bad” lie in the remaining clusters. Real data however might contain more than one class of ”good” data in addition to the outliers group. In this paper the standard method will be extended and applied sequentially. That is, the method is applied on the data to identify a ”good” group in the data, then remove this group and apply the method again to get the next ”good” group from the rest of the data, and so on until all the observations are classified into their groups and one of these group is the outliers class. The main problem, in this case, is that the first ”good” data may only contain a small portion of the observations. This matter will be discussed in detail in this work and some applications on different data are presented.
SourceInternational Conference of Mathematical Sciences
- Makale Koleksiyonu 
The following license files are associated with this item: