The most difficult aspect of conducting research, any research, is the discipline required for data collection and analysis. Designing an experiment is pretty straightforward, and analyzing the information gained is determined by the way the experiment was designed in the first place. Our experiment, testing two different amounts of Basil leaves plus a control, for controlling Varroa mites in a beehive, is a fairly simple setup, but it explores the basics of fundamental research. We used 10 colonies for each of the three treatments (control, half and full treatment), for a total of 30 colonies at the beginning of the experiment.
All 30 colonies were established from packages and queens from a known source and genetic background, fed, medicated and treated in the same manner starting May 15. A month later, baseline readings were taken (June 15), and once a month thereafter, samples of bees (100-200) were taken from the center of the brood nest and stored in alcohol until they could be counted both the bees collected and any Varroa mites adhering to those bees were tallied. The hypothesis was that if Basil leaves did indeed control Varroa mites in a colony, then colonies not treated at all would have the most mites per sampling period and overall, compared to those colonies with a diluted treatment (1/2 cup), which, in turn, would have more mites than those receiving the full treatment of Basil leaves (full cup). Our samples were taken on the 15th of each month, and the colonies receiving the Basil leaves were treated right after the bees were collected. Samples were taken for six months, ending the middle of November. Next, those samples were counted (Please refer to our data collection sheet.) We’ll use the August 15 collection as an example. The check mark by the colony number indicates a sample was taken on the date indicated. Bees are not counted in the field, nor are mites, but any out-of-the-ordinary factor is noted on the sheet. Colony #3 died since the last collection. The bottles the bees were collected in are labeled with the colony number and date of collection for future counting.
Later, the bottles and data sheets are pulled out; the bees and mites in each bottle are counted and the numbers written in the appropriate spaces. Since we randomized our treatments, colony #1 and colony #2 probably aren’t in the same treatment (although randomizing first and numbering later would solve this), so we need to put like colony information with like. That is, make a new sheet that gathers all the information from a single treatment together. This is called a Data Assembly Sheet.
Of course, we could have done this in the first place, or, using any number of spread-sheet computer programs, let the machine do the work. The method described here is over-simplified to make sure things are clearly explained. At the bottom of this second sheet, notice the “total” and “average” rows. These are important numbers, so we need to calculate them carefully. The “average” is the total number collected, divided by the number of colonies we collected from. In the example from July 15 1,365 bees were collected from 10 colonies 1,365¸10=136.5, which is the average number of bees collected per sample for this treatment.
We do these calculations for each treatment, so that when completed, we will have all of our information on three charts, one chart for each treatment. Then we can begin our analysis.
A caution is in order here. Do not be swayed by just your season-long observations. No matter what you are testing, and no matter what your initial thoughts were, rely on the numbers to tell you what’s going on. Another researcher needs to know the techniques you used in order to duplicate the experiment. They probably won’t have the same preconceived notions (or they may be just the opposite) as you did. Observations are important and should be noted, but if you can’t consistently and accurately measure the event it won’t help you prove or disprove your hypothesis.
Next, calculate the mite load, which is simply the number of mites divided by the number of bees for each sample. For example, for colony #2, on August 15, 179 bees and 64 mites were collected. Divide 64 by 179 to find .36 mites per bee in that sample. Also, calculate the total mite load for the treatment for a sampling date. Again, look at the Assembly Sheet for July 15. On that date, a total of 1,365 bees were collected from all 10 colonies, and 326 mites were collected. Divide 326 by 1,365 to find the average of .24 mites per bee per treatment on July 15. Do this for every sampling period, for every treatment. But first, make another chart.
A Summary Chart is strictly that. It is made up of all of the mites per bee calculations from your Data Assembly Sheet. When complete, it shows the results of each colony and each treatment for the duration of your experiment.
You may have several Data Assembly and data summary charts from your experiment. You will have one for each aspect you are measuring, and you may, actually should, measure more than just mites per bee when testing your Basil leaves.
For instance, what about colony weight (a measure of honey production), or brood production (a measure of the queen and Varroa infestation), or incidence of other problems arising during the season (Diseases such as chalkbrood, queenlessness, colony death and the like may be measurable if they occur with regularity)?
There are fairly simple statistical tests that can tell if honey production and mites per bee are related. The question then is: Does honey production in a colony go down as the mites per bee ratio goes up? We’re not going to examine these secondary questions, but you can see where this type of research can lead.
The Summary Chart is just that, only a summary. There are a multitude of statistical analyses that could be run on this data to fine-tune your findings. However, just looking at the chart will tell you if you’ve got something worth further investigation. Often, the numbers alone are difficult to use or interpret. It helps to make a graph of the data from the summary chart, or from any of the data you’ve gathered.
There are several ways to look at the numbers you have worked so hard to gather, but we’ll focus on only three. We didn’t include all of the numbers on our data collection, assembly or summary sheets, but we do have others for these graphs. The first is a look at how a treatment fared over the course of the experiment. Look at treatment 2 from the summary chart, using a bar graph. The graph shows the Varroa story over the life of the experiment for the colonies in that treatment. Remember though, these are averages, not actual numbers.
The second bar graph looks at each treatment at one point in time, (September 15). It tells an interesting story, but only part of it. This last graph shows the average mite load for all treatments for the entire experiment. It tells the whole story of what went on during your experiment. , according to the data you collected. It does not prove Basil leaves control Varroa mites, nor does it prove otherwise. More tests and statistics are needed.
There is one more caution to consider. Your findings are only as good as the design of your experiment and the quality of the data you collect. The design is fairly straight foreward, but do not be tempted to influence the outcome of your experiment to favor your hypothesis, just so you look good. There’s absolutely nothing wrong with being wrong, and proving it. Scientists do it all the time, though they seldom publish those results. You don’t need to either.
However, when you have enough information, collected in a way that is useful, you can approach someone with your hypothesis, your data and your analysis. You may have enough to make some claims about your hypothesis, or there may be absolutely nothing to your idea. The point is, you won’t know unless you pursue the type of research laid out here. Without it, you don’t have a leg to stand on. With it, you may have a beehive full of silver bullets.