Prediction Bias and Response Surface Curvature
Abstract:Many functions used for forestry applications have response surfaces that have curvature. Depending on the degree of curvature, predictions made with these functions can be biased when the inputs to these functions have unbiased random errors. An approximation is presented for calculating the bias in prediction when there are unbiased but random errors in the inputs of the function. Examples of some fairly simple functions used in forestry are shown to provide highly biased predictions in typical applications. For. Sci. 37(3):755-765.
Keywords: Errors in predictors
Document Type: Journal Article
Affiliations: Associate Professor of Forest Biometrics, 110 Mumford Hall, Department of Forestry, 1301 W. Gregory Drive, University of Illinois, Urbana, IL 61801
Publication date: August 1, 1991
- Membership Information
- ingentaconnect is not responsible for the content or availability of external websites