Abstract
NIR news
Volume 20 Issue 5, Pages 16–17 (2009)
doi: 10.1255/nirn.1139
Chemometric Space: Regression statistics in discriminant analysis
Tom Fearn
Department of Statistical Science, University College London, Gower Street, London WC1E 6BT, UK
A commonly used approach to two-group discriminant analysis is to regress an indicator variable that codes for group membership on explanatory variables such as NIR spectra. Having used a regression to derive the classification rule, it is tempting to quote the usual regression statistics: SEC, SEP, R2 and so on. However, these statistics are of limited use in the context of classification, and can be downright misleading.
Keywords:
Full-text article (201 kB) (subscribers only)
Buy article on-line for £12 (get immediate access)
RSS FeedPermalink: http://dx.doi.org/10.1255/nirn.1139
QR Code (what is this?):
Alerting Services
Our Table of Contents Alerting Service will keep you up-to-date with the latest research published in our journals.
You can also follow our journals on Twitter or subscribe to their RSS feeds. 
Subscriptions
Discover the benefits of subscribing to our periodicals
- Quality Science
- Fair Pricing
- Important Research
- Flexible Subscriptions
New Books
New Series of Focused Books in Print and E-Reader Formats
“If you’re going to experiment, then it is always worth doing it properly” writes Tom Fearn in this introduction to Design of Experiments.
find out more
This book provides an overview of the deployment of NIR analysers onto harvesting machinery to give real-time, point-of-cropping data.
find out more
Sample Copy of NIR news
