NON-LINEAR FACTORS IN EMPIRICAL DATA,
Abstract
Five empirical correlation matrices having simplex form, and all requiring no less than two linear common factors to explain them, are each supplied with an alternative nonlinear explanation. For each a single underlying continuum or factor is postulated, the empirical variables being non-linearly but essentially monotonically related thereto. Psychological, anthropometric, or sociological plausibility of these non-linear descriptions is good. The aim was to suggest that non-linearities may occur in real data and that methods for their analysis must be considered. There is no advocacy of abolition of linear methods where these suffice. The non-linear approach discussed here is intended as an addition to, not a substraction from, the scientifically parsimonious linear methods that have served us so well for so long. (Author)
Document Details
- Document Type
- Technical Report
- Publication Date
- Jan 01, 1965
- Accession Number
- AD0623023
Entities
People
- W. A. Gibson
Organizations
- Queens College