1980
DOI: 10.1080/00401706.1980.10486166
|View full text |Cite
|
Sign up to set email alerts
|

Bias in Wilks' A in Stepwise Discriminant Analysis

Help me understand this report

Search citation statements

Order By: Relevance

Paper Sections

Select...
1
1
1
1

Citation Types

2
11
0

Year Published

1986
1986
2015
2015

Publication Types

Select...
7
2

Relationship

0
9

Authors

Journals

citations
Cited by 18 publications
(13 citation statements)
references
References 19 publications
2
11
0
Order By: Relevance
“…However, the possibilities of overinclusion and the selection of an entirely spurious 'best subset' were investigated. If, for instance, too many variables were included, some of which did not contribute to separation, or if none of the included expectancies were actually good discriminators despite apparently acceptable levels of statistical significance, then the solution would not likely be stable, i.e., a different 'best subset' would emerge if the study were repeated (Rencher and Larson 1980). The stability of the solution shown in table 1 was hence investigated by the following cross-validation procedures.…”
Section: Ruling Out Biasmentioning
confidence: 99%
“…However, the possibilities of overinclusion and the selection of an entirely spurious 'best subset' were investigated. If, for instance, too many variables were included, some of which did not contribute to separation, or if none of the included expectancies were actually good discriminators despite apparently acceptable levels of statistical significance, then the solution would not likely be stable, i.e., a different 'best subset' would emerge if the study were repeated (Rencher and Larson 1980). The stability of the solution shown in table 1 was hence investigated by the following cross-validation procedures.…”
Section: Ruling Out Biasmentioning
confidence: 99%
“…7 However, Rencher and Larson [24] have shown that stepwise methods have a high risk of selecting spurious variables, unless the sample size is very large. For this reason we also have estimated 5 This assumption means that the conditional probability P t cannot depend on contemporaneous values of the components of the vector of explanatory variables.…”
Section: Model and Resultsmentioning
confidence: 99%
“…Consequently, selecting a best subset of h values before performing discrimination is generally an attractive classification strategy. We implemented this strategy by invoking the Wilks-Lambda analysis (Rencher & Larson, 1980). The second problem was concerned with the covariance matrices which were not equal or heteroscedastic.…”
Section: Discriminant Analysis (Da)mentioning
confidence: 99%