EXPERIMENTAL DESIGNS OF FIELD EXPERIMENTS IN SINGAPORE : A CRITICAL REVIEW

Abstract

D. A. Preece has indicated from his experience that badly designed, unsatisfactorily run and unsoundly analysed experiments are common in Third-World countries. The abstract mathematical approach used in statistical textbooks is partly to be blamed. This statistical ritual with a 'cookbook' approach to both design and analysis has taken over from scientific thinking. A review of recent issues of the Singapore Journal of Primary Industries shows that the above phenomena are also prevalent in Singapore, particularly in the field experiments. Multiple comparison tests are often misused in many experiments and randomised complete block design seems to be used in every field experiment, regardless of the nature of site. In this academic exercise, the planning, design and analysis of experiments are discussed, with more emphasis being placed on the latter two aspects. Examples are given to illustrate both the appropriate and inappropriate uses of multiple comparison tests. For those examples in which these tests are used erroneously, attempts are made to present a more meaningful statistical procedure. Moreover, some alternatives to randomised block design are also proposed and illustrated with examples. The examples used are, as far as possible, experiments done in Singapore. The alternatives to randomised block design make use of the analysis of covariance technique which aim to eliminate the uncontrollable environmental variations that may not be effectively eliminated by blocking. They have proved to be as efficient as, if not more than, the randomised block design, except in one of the examples.