Sample size and class variability in the choice of a method of discriminant analysis
- 10 May 1994
- journal article
- research article
- Published by Informa UK Limited in International Journal of Remote Sensing
- Vol. 15 (7), 1551-1555
- https://doi.org/10.1080/01431169408954189
Abstract
The choice between linear and quadratic discriminant analysis in a simulation, (where DN values are normally distributed within classes and the classes are of known prior probability), is shown to depend on the size of the training sample and the extent to which the within-class variance-covariance matrices difTer from each other.Keywords
This publication has 3 references indexed in Scilit:
- Statistical problems in the discrimination of land cover from satellite images: a case study in lowland BritainInternational Journal of Remote Sensing, 1992
- Maximum likelihood classification, optimal or problematic? A comparison with the nearest neighbour classificationInternational Journal of Remote Sensing, 1987
- Discrimination problems for satellite imagesInternational Journal of Remote Sensing, 1987