Is regression what you really want? |
| |
Authors: | R. Webster |
| |
Affiliation: | Rothamsted Experimental Station, Harpenden, Herts AI.5 2JQ, UK. |
| |
Abstract: | Abstract. Regression is frequently abused in soil research. Its proper use is for statistical prediction. It may also be used to calculate equations for calibration. A regression equation may be used to express a functional relation between two soil variables that are thought to be related by some simple mathematieal law but only where one of the variables is known exactly. In most other circumstances regression is inappropriate. Where departures from a functional relation are due to errors of measurement and sampling fluctuation it should be replaced by a structural analysis to find the best equation. Where the underlying relation is truly bivariate it should be described as such. |
| |
Keywords: | |
|
|