9 Logistic Regression
The purpose of many research projects is to assess relationships among a set of
variables and regression techniques often used as statistical analysis tools in the
study of such relationships. Research designs may be classified as experimental
or observational. Regression analyses are applicable to both types; yet the
confidence one has in the results of a study can vary with the research type. In
most cases, one variable is usually taken to be the response or dependent vari-
able, that is, a variable to be predicted from or explained by other variables.
The other variables are calledpredictors,explanatory variables,orindependent
variables. Choosing an appropriate model and analytical technique depends on
the type of variable under investigation. Methods and regression models of
Chapter 8 deal with cases where the dependent variable of interest is a con-
tinous variable which we assume, perhaps after an appropriate transformation,
to be normally distributed. In Chapter 8 we considered cases with one inde-
pendent variable (simple regression) and cases with several independent vari-
ables or covariates (multiple regression). However, in a variety of other applica-
tions, the dependent variable of interest is not on a continuous scale; it may
have only two possible outcomes and therefore can be represented by an indi-
cator variable taking on values 0 and 1. Consider, for example, an analysis of
whether or not business firms have a day-care facility; the corresponding inde-
pendent variable, for example, is the number of female employees. The depen-
dent variable in this study was defined to have two possible outcomes: (1) the
firm has a day-care facility, and (2) the firm does not have a day-care facility,
which may be coded as 1 and 0, respectively. As another example, consider a
study of drug use among middle school students as a function of gender, age,
family structure (e.g., who is the head of household), and family income. In this
314
