Understanding What Practices Work Best

Evaluating

Each grower should develop an understanding of what practices work best on a particular farm. In some cases, research may have been conducted under different conditions or in another area and may not apply in a particular farmer’s field.

However, simple comparisons in the field can help a farmer understand the value and the risk of different practices. These comparisons will show if an input (inoculant, fungicide or herbicide) contributes to yield and profitability. To evaluate these inputs under the real conditions of a particular field, a farmer must avoid biased tests and replicate the treatments to have confidence in the results.

On one field, apply inputs in one place and do not apply inputs in another, in order to compare the results. To avoid bias, select fields and areas within the field where the problem exists. In other words, herbicide should be tested in an area that has a history of weed problems.

Replicating results takes more effort and planning, but is important because the soil type, drainage and pest pressure can vary across a field. Repeat the treatments in multiple locations to account for that variation and avoid choosing an area that might have more pressure than others. For example, if the edge of the field has many more weeds than the rest of the field, testing there might not accurately demonstrate the overall effect of the herbicide.

While putting treatments out in multiple strips in a field is a good start, the preferred method is to replicate the treatments randomly. The figure below illustrates a good example of a randomized trial comparing one input. If possible, it is good to replicate at least four times for each treatment.

After the appropriate data are recorded, it is then important to analyze the data using appropriate statistical procedures. All treatments (such as planting an improved variety) need to have an equal chance of being placed at random somewhere in the experimental area. Likewise, all treatments need to be replicated a reasonable number of times (in most agricultural and pest management settings four replicates are sufficient). One should never draw major conclusions from a single experiment. Experiments should be repeated over space (locations) and time (growing cycles) to characterize consistently in response to variables. For example, many breeding programs require that a group of varieties be compared for three years at five locations before release.

Example of a replicated and randomized trial area for two inputs. 1 = Input #1, 2 = Input #2, 3 = Input 2 + 3, 4 = No Inputs or Untreated