Plant Growth-Analysis: Further Applications of a Recent Curve-Fitting Program

Publisher Website
Google Scholar
Abstract
Two points emphasized are the use of regression techniques in plant growth-analysis: the selection of the basic equations necessarily controls the behavior of the derived quantities; small improvements in fit bring large benefits later. For example, the regression applied to the whole growth period is significant at P < 0.001 and the regression accounts for a very high percentage of the total variation. Yet, improvements are both possible and worth while. A complex series of data could be approached by this segmentation process if in each segment there was enough information for satisfactory regression. The separation is subjective and the experimenter''s division would be open to debate. For example, in the present set of data, a short period of true exponential growth (a plateau in R) might be recognized between days 43 and 64, leading to an analysis in 3 segments. Alternatively, medium-scale deviations from the otherwise smooth trends might be pursued individually in an attempt to link growth to some fluctuating environmental condition. In each case a combination of subjectively divided data and objectively determined regression models would provide the basis for deriving the quantities needed. Work in progress involves the application of splined regressions (Cox, 1972), using both subjectively and objectively determined knots.