Descriptive studies tend to be simpler and easier to conduct than analytical, experimental, or quasi-experimental studies, but they are nonetheless quite important. Descriptive studies can provide the background from which analytical, experimental, or quasi-experimental studies emerge. Descriptive studies help to generate hypotheses, as opposed to testing them.
Typical statistics are measures of dispersion and central tendency. The most familiar measures of dispersion are the variance and standard deviation. Common measures of central tendency are the mean, median, and mode.
You do, however, sometimes see measures of association in descriptive studies. The Pearson product-moment correlation coefficient is a very popular measure of association, but there are many others.
Measures of dispersion, central tendency, and association provide important clues about what to expect if you select one or more of the included variables to define an analytical, experimental, or quasi-experimental study. For example:
- Variables with greater dispersion tend to be more desirable, statistically speaking, than variables with lesser dispersion. For example:
- Suppose you have developed some sort of screening test - say - a test that you think may indicate a person's chance of contracting AIDS given exposure to the HIV virus.
- If you test a bunch of people and they all test about the same, this test probably isn't going to have much predictive power, is it?
- On the other hand, if you test a bunch of people and see a wide range of scores, you are well on your way to testing the hypothesis that more people who score high on this test contract AIDS than do people who score low on this test (everything else equal), are you not?
- Measures of central tendency are useful in two ways:
- They indicate what might be considered a "normal" score on your measure (which might be defined by the mean, median, or mode, depending on the specific measure), and:
- They indicate the "normality" of your data. If the mean, median, and mode are all about the same, your data is more likely to be appropriate for parametric analysis than if they are widely different. Wide variation in these measures may also indicate that you need to take a closer look at your sample selection process or increase your sample size. (This is an advanced topic.)
- Measure of association also help determine which variables may be useful for further studies. For example, if you have two variables which correlate very highly (assuming they show a similar pattern of correlation with other variables in the dataset), then you probably can eliminate one of them as they seem to be measuring pretty much the same thing.
Return to the research flowchart.