HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS		QUICK SEARCH:   [advanced] Author: Keyword(s): Year:  Vol:  Page:

This Article Full Text (PDF) Alert me when this article is cited Alert me if a correction is posted Services Similar articles in this journal Similar articles in PubMed Alert me to new issues of the journal Download to citation manager Citing Articles Citing Articles via Google Scholar Google Scholar Articles by Cox, D. F. Search for Related Content PubMed PubMed Citation Articles by Cox, D. F.

Design and Analysis in Nutrional and Physiological Experimentation¹

Department of Statistics, Iowa State University, Ames 50011

ABSTRACT

Conclusions: The concepts and suggestions discussed here are not new but come from the basic texts on statistical methods (4). These texts often are cited in support of methods that they do not encourage or condone. My concern is that too much of the data analysis reported in recent journal articles is sometimes simply incorrect and more often unnecessarily obscure and cumbersome. The most serious error that I find is the failure to understand what is meant by an experimental unit and, consequently, to conduct unreplicated experiments. This problem often becomes acute when units are measured repeatedly. Reported research could be easily clarified and simplified by choosing more appropriate scales to record observations. The multiple-range tests seem to have evolved as the main evaluative tool for examining experimental results. They are now used almost universally and in situations in which they were never intended to apply. Often, much simpler and more direct methods would be more appropriate, sufficient, and clearer. Polynomial regression in nonlinear situations is sometimes the only possibility, but indiscriminate use of higher-order polynomials seldom clarifies much of the underlying mechanisms and can be extremely misleading.

What one sees in the uses of statistical methods today is a consequence of attempts to force investigators not merely to describe their results but to evaluate quantitatively the differences that they find in the light of the experimental error in the measurements they make. The emphasis in basic textbooks and in most teaching is heavily toward testing the significance of results. What is needed is a return to a place where a clear exposition and description of the mechanism producing the differences in experiments are given the central role they deserve without neglecting the necessity of examining the strength of the evidence for a particular point of view.

¹ Journal Paper No. J-9565 of the Iowa Agricultural and Home Economics Experiment Station, Ames, Project No. 101. Invited paper at the 74th Annual Meeting of American Dairy Science Association, June 25 to 27, 1979, Utah State University, Logan.