Statistical control of variability in coconut experiments

Abstract

High variability in plot yields arising from changes in the number of bearing palms and their bearing age is a feature of experiments with young coconuts. As correction by pre-experimental calibrating variates is inapplicable to this situation, the use of certain intra-experimental calibrating variates is suggested. Empirical formulae defining the curvilinear influence of bearing-palm number and age on yield have been postulated. The use of multiple covariance analysis as a means of testing the validity of the empirical relationships and of detecting the suitability of calibrating variates has been explained. The results of the application of the above methods in a fertilizer experiment with young coconut palms have been used to illustrate the extent of the reduction in experimental error and the effect of such adjustments on the interpretation of data. The importance of such forms of statistical control both from the point of view of increasing experimental precision and of ensuring correct interpretation of data is emphasized.