|
The answer to question one governs whether you use (a) a test of differences or (b) calculate a correlation coefficient or some other measure of the degree to which variables are associated, such as chi-square. In answer to question two, data are interval when you have measured your dependent variable on a scale that enables you to assume that the gap between any two adjacent scores is always the same.
For instance, we know that the gap between 10 degrees and 20 degrees Celsius is the same as the gap between 50 and 60 degrees Celsius. The difference in temperatures is the same wherever we take temperatures that are 10 degrees Celsius apart. However, is the gap between an IQ of 105 and one of 115 the same as the gap between IQ 145 and IQ 155? That is, does a gap of 10 IQ points measure the same amount of intelligence wherever it occurs? If not, then IQ scores are not measured on an interval scale. Indeed, many of the scales used in psychological measurement probably do not have equal intervals between adjacent points.
Where we are unable to assume equal intervals between points on the scale, we can often at least rank order the numbers. That is, we can say that an IQ of 155 is greater than one of 145, and less than one of 165, even if we cannot assume that it is equidistant between them. When this is the most that we can do, we call our data ordinal.
There are occasions, however, when we cannot even rank order our data. When this occurs - when our data are categories composed of frequencies - then they are nominal. For example, in the cheese and nightmare experiment in Chapter 9 of the book, our data simply fell into the categories nightmare or no nightmare. We have no idea whether those in the experimental condition experienced nightmares that were more nightmarish than those in the control condition. All we can say about any given participant was whether they fell into the category nightmare or no nightmare. Such data - counts of individuals who fall into particular categories - are nominal.
Different tests are available for these different types of data. In order to know which test to use, therefore, you need to be able to recognize which type of data you have.
|
