Premium
Analysis of a Randomized Block Design with Unequal Subclass Numbers
Author(s) -
Piepho HansPeter
Publication year - 1997
Publication title -
agronomy journal
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.752
H-Index - 131
eISSN - 1435-0645
pISSN - 0002-1962
DOI - 10.2134/agronj1997.00021962008900050002x
Subject(s) - subclass , block (permutation group theory) , mathematics , randomized block design , statistics , variable (mathematics) , algorithm , computer science , combinatorics , medicine , antibody , mathematical analysis , immunology
For some types of agricultural experiments, there are many observations per experimental unit; i.e., the subclass number is greater than one. In this article, the focus is on the randomized complete block design. Practitioners are often faced with unequal subclass numbers due to unpredicted experimental problems. The present paper discusses the statistical analysis of such unbalanced data, where subclass numbers vary independently of treatment variables and the response variable is not influenced by subclass number. In an example using the SAS statistical package to analyze data from a randomized complete block greenhouse experiment, an analysis of cell means compared favorably with a more elaborate Satterthwaite method. Monte Carlo simulations yielded similar results for both methods. It is concluded that, in practice, the cell means analysis is preferable for its simplicity.