Studies on Data Variability and the Use of Polynomials to Describe Plant Growth

SUMMARY The question of which degree of polynomial should be fitted to plant data, for the purposes of growth analysis, has been considered in relation to the variability of the populations sampled. By means of actual and semi‐artificial data, it was shown that the variability of the samples taken at each harvest had a profound effect on the results of the tests of significance, when the relationship was fitted using individual observations. Significance tests performed on data of low variability resulted in over fitting, i.e. unrealistically high degrees were indicated as being most appropriate. Conversely, data of high variability were adequately fitted by lower order polynomials. In view of this, it is suggested that growth data should be fitted using harvest mean values, since in this case, the test of adequacy of fit is independent of the underlying population variability.