ve (Pfaffenberger and Patterson, 1991, pp. 864-865). The logistic curve is based on the exponential of a function, and is shaped like the letter "s." This transformation, referred to as "logit transformation," is, in effect, a transformation of the conditional probabilities of a dichotomous variable (Dwyer, 1983, p. 447).
The second problem that arises in relation to regression analysis with respect to dichotomous dependent variables is that the relationship between the independent and dependent variables is not additive (Dwyer, 1983, p. 447). A multiplicative model is more appropriate for use with such variables (Dwyer, 1983, p. 447). Although the multiplicative model is not linear in character, the character can be made to be linear through logit transformation (Dwyer, 1983, p. 447).
Dwyer (1983, pp. 447-453) provided an illustration of the type of problem that is susceptible to solution through logit transformation. For this illustration, assume that data were available indicating the highest level of formal educational attainment for both a group of fathers and the sons of those fathers. If one wished to predict the highest level of formal educational attainment of a son based on the highest level of educational attainment of his father, such a prediction could be derived through the solution of the following regression equation:
Y = c0 + c1X + u, in which Y represents the highest level of formal educational attainment of the son, and in which X represents the highest level of educational attainment of the father.
Suppose for purposes of illustration, however, that the data available on the formal educational attainment of the sons was incomplete, and that the only facts that could be supported from the data were that the subjects either graduated from college or did not so graduate. The dependent variable in this instance would, thus, be d...