Hostname: page-component-586b7cd67f-g8jcs Total loading time: 0 Render date: 2024-11-27T19:48:57.715Z Has data issue: false hasContentIssue false
  • English
  • Français

Data Analysis in Agricultural Experimentation. I. Contrasts of Interest

Published online by Cambridge University Press:  03 October 2008

S. C. Pearce
S. C. Pearce
Affiliation:
Applied Statistics Research Unit, University of Kent at Canterbury, England
Get access Rights & Permissions [Opens in a new window]

Summary

An experiment has its origin in the need to find answers to stated questions. From the start great care is given to its correct conduct (e.g. the application of treatments and the method of recording) as well as to statistical design, always with the original questions in mind. The analysis of its data is the climax of a long process and the questions to be answered must dominate all else. It is not enough to feed data into a computer package in the hope that it will provide an automated path to a true interpretation.

Where the treatments have been chosen with care to answer specific questions, the statistical way of designating purpose is to declare ‘contrasts of interest’, each corresponding to a degree of freedom between treatments. They derive solely from the reasoning behind the selection of treatments. If possible the questions posed should be equal in number to the degrees of freedom and should admit of separate study because no one can give a single answer to several diverse questions.

This paper shows how to define a contrast of interest and how to isolate it in the analysis of variance. Attention is given both to its contribution to the treatment sum of squares and to its variance (i.e. the precision with which it is estimated). Independence of estimation is also considered. Algebraic formulae are given for a restricted though important range of designs, which includes those that are completely randomized, in randomized blocks or in Latin squares, all treatments having the same replication. The methods can, however, be generalized to cover all designs. With these formulae it is possible both to test the existence of an interesting effect and to set confidence limits round its estimated value.

Type
Research Article
Information
Experimental Agriculture , Volume 28 , Issue 3 , July 1992 , pp. 245 - 253
Copyright
Copyright © Cambridge University Press 1992

Access options

Get access to the full version of this content by using one of the access options below. (Log in options will check for institutional or personal access. Content may require purchase if you do not have access.)

References

Chew, V. (1980). Testing differences among means: correct interpretation and some alternatives. HortScience 15: 467–470.CrossRefGoogle Scholar
Gill, J. L. (1973). Current status of multiple comparisons of means in designed experiments. Journal of Dairy Science 56: 973–977.CrossRefGoogle Scholar
Mead, R. (1988). The Design of Experiments. Cambridge: Cambridge University Press.Google Scholar
Pearce, S. C. (1983). The Agricultural Field Experiments: A Statistical Examination of Theory and Practice. Chichester: John Wiley and Sons.Google Scholar
Pearce, S. C., Clarke, G. M., Dyke, G. V. & Kempson, R. E. (1988). A Manual of Crop Experimentation. London: Charles Griffin, and New York: Oxford University Press.Google Scholar
Tharp, W. H., Wadleigh, C. H. & Barker, H. D. (1941). Some problems in handling and interpreting plant disease data in complex factorial designs. Phytopathology 31: 26–48.Google Scholar
Tukey, J. W. (1991). The philosophy of multiple comparisons. Statistical Science 6: 100–116.CrossRefGoogle Scholar