In the agricultural realm we have access to many types of data. One of the most difficult things in science is being able to determine how much data is needed to make a reasonable conclusion about the data, and whether the data is meaningful at a larger scale. In particular, there has been much discussion about the value of small research plots vs large strip trials. Both have advantages and disadvantages, which can be leveraged depending on the study.
Small plot research can be conducted on a relatively small area, easily replicated, and allows for multiple treatments to be tested simultaneously (think nitrogen x spacing x fungicide timing for example). Limitations in plot size can result in larger than typical variability, especially with corn research. This is because a small amount of ear loss at harvest can have a fairly large effect on overall plot yield, and the number of plants in the plot is small. Thus, small differences in the absolute number of plants in the plot and ears harvested can result in more variability in yield from plot to plot. In the past, many also discounted small plot research due to perceived edge effects from and alleys influencing results. However, at least with fungicide work, research indicates that alleys and edge effects do not influence overall results (Vincelli and Lee 2015).
Contrast this with larger scale studies such as strip trials. These studies do not suffer from as many issues yield-related variability, at least regarding corn, because plots are much larger and therefore grain samples are less influenced by ear loss and limited population. The downside is that due to the plot size, strip trials are often limited to a small number of treatments. Consequently, the same trial needs to be conducted across many sites and years before enough observations are made to draw a meaningful conclusion.
That brings me to my last point. Some people, for whatever reason, completely discount small plot research results. Often, many of the aforementioned claims are mentioned in this argument. In the end, is there really any difference in results? Researchers at Iowa State recently published a study that examined small plot and strip plot fungicide trials conducted in Iowa from 2008-2015 (Kandel et al 2015). A total of 230 strip plot trials and 49 small plot trials were included in the analysis. Based on their analysis, the researchers found that the yield responses for the various treatments were similar. However, the data from small plot trials were slightly more variable. For example, to detect a given yield response, a split plot trial would need three treatment replications per field and 12 locations, whereas a small plot trial would need seven replications at each site and 12 locations. Thus, based on their work, small plot data appear to be more variable than larger scale trials, but the overall responses are not different, only the variability. Increasing the amount of replication at a site or overall observations can overcome this variability. Changing the level of significance used to call something significantly different may help as well. Their results show that small plot data show the same results as the large plot data, but might need a little more replication to detect a difference when compared to strip trials. Both small plot and large plot data are useful for agriculture, and both have their advantages and disadvantages.
Lastly, I will leave you with a link to a great little article from the University of Nebraska called “Field Studies: What do you mean 5 bushels per acre is not significant?” In this article the authors do a great job of discussing research and terms such as significance and variability. This is a good read heading into meeting season and Crop School.