VARIABILITY IN ENZYME-LINKED IMMUNOSORBENT ASSAYS AND CONTROL OF EXPERIMENTAL ERROR BY USE OF EXPERIMENTAL-DESIGNS

Seven uniformity trials were done on Nunc-Immuno IF and Dynatech Immul on 1 ''U'' microplates with two monoclonal triple-antibody sandwich en zyme-linked immunosorbent assays to determine if an experimental desig n was necessary to control plate variability. A single uniform extract of either oat (cv. Clintland 64) tissue infected with barley yellow d warf luteovirus (BYDV-PAV-IL) or soybean (cv. Williams 82) tissue infe cted with soybean mosaic potyvirus (SMV-G5) was prepared, and samples were placed in the wells of eight microplates used for each trial. For whole plates, mean absorbances varied within trials and coefficients of variability ranged from 3.8 to 20.3%. Seventeen of the 56 plates te sted had one to five wells with absorbances that were outliers from th e normal distribution. Nonrandomized designs of two replications of 48 treatments were also assigned to each plate. When replications of a t reatment were paired in either rows or columns, differences between tr eatment means were confounded with row and column differences, indicat ing the need for an experimental design on microplates. Both the rando mized complete block design and the alpha (0,1) one-restrictional reso lvable incomplete block design allowed for more precision than the com pletely randomized design. The smaller, incomplete blocks of the alpha design afforded slightly more precision than the larger, complete blo cks of the randomized complete block design.