CHAPTER 18 A Yes-or-No Proposition: Logistic Regression 257

The predictors can be quantitative, such as age or weight. They can also be

categorical, like gender or treatment group. You will need to make decisions

about how to recode these variables to enter them into the regression model.

See Chapter 17, where we describe how to set up categorical predictor

variables.

3.

Tell your software which variables are the predictors and which is the

outcome.

Depending on the software, you may do this by typing the variable names,

or by selecting the variables from a menu or list.

4.

Request the optional output from the software if available which may

include:

• ^{A summary of information about the variables, goodness-of-fit measures,}

and a graph of the fitted logistic curve

• ^{A table of regression coefficients, including odds ratios (ORs) and their}

95 percent confidence intervals (CIs)

• ^{Predicted probabilities of getting the outcome for each individual, and a}

classification table of observed outcomes versus predicted outcomes

• ^{Measures of prediction accuracy, which include overall accuracy, sensitiv-}

ity, and specificity, as well as a Receiver Operator Characteristics (ROC)

curve

5.

Execute the model in the software.

Obtain the output you requested, and interpret the resulting model.

Interpreting the Output of Logistic

Regression

Figure 18-4 shows two kinds of statistical software output from a logistic regres-

sion model produced from the data in Table 18-1. The output presented is not from

a particular command in a particular software. Instead, typical output for the dif-

ferent items is presented to enable us to cover many different potential

scenarios.