Approaching Data

So, you've run your experiment, collected all the results, and now you've got a huge pile of data in front of you. To make matters worse, you've got a lab report that's due pronto. Don't panic. It is survivable.

At this point you may be feeling an almost irresistable urge to dive in and start attacking the data: running statistics, plotting graphs, constructing tables. Whoa! Slow down. Head down that path too quickly and you'll end up making more work for yourself than is absolutely necessary. It is very easy to get lost in the (seemingly) big issues of what sort of statistical test to run (e.g., t-test, Chi-squared, etc.), or what kind of graph to plot (e.g., line graph, histogram, etc.). Before going this route take a mental step back and try looking at the big picture.

What "Big Picture?" I hear you ask. Fortunately, I hear your cry and have some answers for you. At the heart of any research design lies a question or questions. It is the questions that you should concentrate on when thinking about the data.

When approaching any collection of data I suggest that you begin by asking what basic questions could be generated from the experimental design that produced the data. Once you begin to formulate questions the data analysis will probably start to fall into place by itself. Often these questions are not terribly complex. In fact, they sometimes seem so simple that they are overlooked.

For instance, let's say you've run an experiment in which you deliver reinforcers to a rat over time. Whether or not the rat is reinforced depends on what stimulus is presented. You record the behaviour of the rat just before and during the presentation of a stimulus.

Just from this information alone you can find the basis of some of the questions to be asked. Notice that reinforcers are being delivered "over time". Whenever you see a protocol in which time is a parameter you should stop and consider just what consequences might result. In this case, you might ask, "How does the rat's behaviour change over time?" Or, even more basically, "Does the rat's behaviour change over time?" If these seem like important questions to you they they will give you a handle on how to approach the data. Right off the top you should be considering some sort of interpretation/presentation of the data that has time as one of the indices of analysis/interpretation. For example, in a graphical format you might show time on the X-axis. Other questions that might suggest themselves to you might include "Do behaviours differ in the pre- and during-stimulus periods?" or "Are behaviours stimulus dependent?"

For whatever data you happen to have I would suggest to you that a good starting point for analysis is to consider what question(s) the experiment was designed to answer. Once you identify the questions you will probably find that making decisions on data analysis and presentation are somewhat more focussed and less haphazard.