Answers are grounded in this paper and cite their sources.
Crop performance trials are conducted in nearly every state as part of the land-grant mission. These trials have been conducted for flue-cured tobacco in the southeastern United States since 1954 and continue today. Even though the trials are conducted in a scientific manner, they are subject to variability caused by management and natural causes like diseases and insects. This article looks at the variability of 3 measurements from the North Carolina Official Flue-Cured Tobacco Variety Trials from 1981 through 2001. Yield, grade index, and dollar value per hectare ($/ha) were examined with the use of a procedure developed for other crops. First the relationship between error and mean was established. For these 3 traits there were no relationships between error and mean. A pooled (average) variance was calculated and then a maximum allowable variance was determined by multiplying the pooled variance by 2. Six out of 112 environments had questionable accuracy for yield; there were 7 questionable environments for grade index, and there were 6 for $/ha. Five questionable environments were common between yield and $/ha and 1 other $/ha environment had high error for yield. Grade index had only 1 common questionable environment with yield and none with $/ha. It is advocated that those who run crop performance trials examine their data for accuracy before publication and choose an alpha level between 0.20 and 0.40 to avoid Type II errors.
No takes yet. Share an insight, caveat, or question.
Daryl T. Bowman (2015) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: