Variables, Categorical

Categorical variables are sets of variables with values assigned to distinct and limited groups or categories. Categorical variables take on values in a set of categories, different from a continuous variable, which takes on a range of values. Categorical variables are also called discrete or nominal variables.

Categorical variables are common in social and behavioral sciences. Groups or categories often [Page 1836]consist of numeric (e.g., female = 1, male = 0) or alphabetic (e.g., female, male) labels, and generally provide information that is not quantitative by nature. In their simplest form, categorical variables can be binary variables with only two options (e.g., “yes” or “no”). In more complex forms, categorical variables can represent a variety of options belonging to the same group or category, such as marital status, religion, or ethnicity. For example, marital status can be described in a variety of ways such as single, married, divorced, or widowed. It is important to note that numeric labels are simply labels and do not indicate one category is more/less or better/worse than the other. Numeric labels are simply codes that allow the researcher to conduct analyses, and do not reflect any rank ordering or quantities. Such codes are referred to as dummy codes, or the quantification of a variable, which allows the researcher to conduct analyses with numeric symbols taking the place of words.

Categorical scales are common in social sciences for measuring attitudes, opinions, and behaviors, and have two primary types of scales: nominal and ordinal levels of measurement. When variables have categories without a natural ordering (serving as labels only), the variable is nominal. Nominal measurement levels consist of only categorical variables that do not have higher or lower status than one another, so the order of listing is irrelevant. The marital status example described earlier, then, is considered a nominal categorical variable. Categorical variables at the nominal level should have no logical order, and be exhaustive and mutually exclusive (each case to be categorized in the nominal measure must fall into only one of the categories). Besides gender, other more commonly used nominal measures in survey research include race, religious affiliation, and political affiliation. Once all cases in the population are examined, cases sharing the same criteria are grouped into the same category and receive the same label (e.g., “female” or “male”).

Categories can also be measured on an ordinal level when they can be naturally ranked (e.g., from lowest to highest), such as size or social class. Ordinal variables are ordered, but the distances between each category are not known. Categorical variables at the ordinal level should also be mutually exclusive, and they should have a logical order and be scaled according to the amount of the identified characteristic. For example, grades can be ordered A, B, C, D, and F, where an individual who earns an A achieves at a higher level than one who earns a B, and so on. As is the case with ordinal level measurement, one should not assume the distance between an A and a B is the same as the distance between a B and a C. Researchers might also use Likert scales, for example, ranging from strongly disagree to strongly agree, indicating more/less agreement with statements. Ordinal categorical variables such as these create qualitative comparisons, which then become meaningful; it makes sense to say that a person who earns an A has achieved at a higher level in a course than a person who earns a B.

...