Premium
The Necessity of Independent Testing of Soil‐Site Equations
Author(s) -
McQuilkin Robert A.
Publication year - 1976
Publication title -
soil science society of america journal
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.836
H-Index - 168
eISSN - 1435-0661
pISSN - 0361-5995
DOI - 10.2136/sssaj1976.03615995004000050044x
Subject(s) - site index , statistics , standard error , mathematics , sampling (signal processing) , index (typography) , regression analysis , geography , forestry , computer science , filter (signal processing) , world wide web , computer vision
Abstract A black oak ( Quercus velutina ) soil‐site study in Missouri has shown that the standard regression statistics ( R 2 , standard error, etc.) of soil‐site equations are not reliable indicators of the accuracy of these equations. Data from 54 plots were used to derive a soil‐site equation that had an R 2 of 0.66 and a standard error of 1.00 m. These statistics plus a graphed comparison of observed vs. predicted site indexes indicated that the equation satisfactorily predicted site index. When the equation was tested with data from 27 additional plots, however, the correlation between the observed and predicted site indexes had an R 2 of only 0.01 and a standard error of 1.81 m. These statistics plus a graphed comparison of observed vs. predicted site indexes showed that the equation was completely unreliable as a site index predictor. Inadequate sampling, incomplete data analysis, or chance correlations may have caused the errors in the equation, but the errors were undetectable until the equation was tested with the independent data.