both scales of measurement, discrete and continuous, even though in most
practical applications, the independent variable under investigation is often on
a continuous scale.
9.1.1 Simple Logistic Regression Model
The usual regression analysis goal, as seen in various sections of Chapter 8, is
to describe the mean of a dependent variableYas a function of a set of pre-
dictor variables. The logistic regression, however, deals with the case where the
basic random variableYof interest is a dichotomous variable taking the value
1 with probabilitypand the value 0 with probabilityð 1 pÞ. Such a random
variable is called apoint-binomialorBernouilli variable, and it has the simple
discrete probability distribution
PrðY¼yÞ¼pyð 1 pÞ^1 y y¼ 0 ; 1
Suppose that for theith individual of a sampleði¼ 1 ; 2 ;...;nÞ,Yiis a Ber-
nouilli variable with
PrðYi¼yiÞ¼piyið 1 piÞ^1 yi yi¼ 0 ; 1
The logistic regression analysis assumes that the relationship betweenpiand
the covariate valuexiof the same person is described by the logistic function
pi¼
1
1 þexp½ðb 0 þb 1 xiÞ
i¼ 1 ; 2 ;...;n;
The basic logistic function is given by
fðzÞ¼
1
1 þez
where, as in this simple regression model,
zi¼b 0 þb 1 xi
or, in the multiple regression model of subsequent sections,
zi¼b 0 þ
Xk
j¼ 1
bjxji
representing an index of combined risk factors. There are two important rea-
sons that make logistic regression popular:
SIMPLE REGRESSION ANALYSIS 317