ERIC Number: ED450146
Record Type: Non-Journal
Publication Date: 2001-Feb
Reference Count: N/A
Linear and Quadratic Rules: What Are They and When Are They Used?
Altman, Daniel R.
Discriminant analysis is a multivariate method of analysis with two purposes: (1) to describe differences among groups; or (2) to classify participants into groups. Either linear or quadratic rules can be used in both descriptive discriminant analysis (DDA) and predictive discriminant analysis (PDA). In both DDA and PDA the researcher wants to use the rules that are the most accurate and the most replicable, but these two considerations often offer competing perspectives about which rule is optimal in a given analysis. The linear rule is used when group covariance matrices are pooled. The linear rule has greater external generalizability because fewer parameters are estimated. The quadratic rule is used when separate group covariance matrices are used. These covariance matrices do not need to be as homogenous. However, the quadratic rule's external generalizability is not as accurate. It is important to use the correct rule in both PDA and DDA. (Contains 10 references.) (Author/SLD)
Descriptors: Multivariate Analysis
Publication Type: Reports - Descriptive; Speeches/Meeting Papers
Education Level: N/A
Authoring Institution: N/A