Abstract

Interim monitoring of accumulating data has been widely used in clinical trials, but it has not received the same attention in agricultural experimentation. The methodology, however, can be a useful tool in agronomic trials designed to find better production techniques or optimal animal treatments at low cost, plus the possible economic advantages resulting from correct early decisions. These sequential procedures for testing hypothesis with available data in successive periods of time dictate termination of the experiment when a significant difference is detected, or otherwise continuation of the experiment to the end of the stipulated time or until all the planned sample size is realized. The statistical cost of repeated testing of part of the same data is a reduction in the significance levels a to the time-related significance levels αjj<α). We apply three methods for this type of analysis, which we illustrate with two examples involving respectively, comparisons of two proportions and two means from normally distributed random variables with unknown variances. The examples show the usefulness and limitations of the proposed methods and also that there can be no absolute rule for choosing the best method of analysis in a particular case. The optimal strategy depends on the specifics of the trial and the investigator's criterion to choose the αj.

Keywords

Stopping rules, interim analysis, agricultural experimentation sequential testing.

Creative Commons License


This work is licensed under a Creative Commons Attribution-Noncommercial-No Derivative Works 4.0 License.

Share

COinS
 
Apr 23rd, 1:30 PM

SEQUENTIAL ANALYSIS OF AGRICULTURAL EXPERIMENTS

Interim monitoring of accumulating data has been widely used in clinical trials, but it has not received the same attention in agricultural experimentation. The methodology, however, can be a useful tool in agronomic trials designed to find better production techniques or optimal animal treatments at low cost, plus the possible economic advantages resulting from correct early decisions. These sequential procedures for testing hypothesis with available data in successive periods of time dictate termination of the experiment when a significant difference is detected, or otherwise continuation of the experiment to the end of the stipulated time or until all the planned sample size is realized. The statistical cost of repeated testing of part of the same data is a reduction in the significance levels a to the time-related significance levels αjj<α). We apply three methods for this type of analysis, which we illustrate with two examples involving respectively, comparisons of two proportions and two means from normally distributed random variables with unknown variances. The examples show the usefulness and limitations of the proposed methods and also that there can be no absolute rule for choosing the best method of analysis in a particular case. The optimal strategy depends on the specifics of the trial and the investigator's criterion to choose the αj.