Regression of a model on real-system output: An invalid test of model validity

Statistical tests of validity of farming systems models may be inappropriate for a number of reasons. A specific example is the F test for zero intercept and unit slope; this test has intuitive appeal, but bias in parameter estimates can lead to rejection of a valid model. It is suggested that descriptive statistics and subjective tests be used to build up confidence in a model as it proceeds through a number of prototypes.