文档介绍：Chapter 7
USING LOGISTIC REGRESSION
Background Information
Logistic regression analysis allows analysts to estimate multiple regression models
when the response being modeled is dichotomous and can be scored 0,1; that is, the
e must be one of two choices. Analysts use logistic regression when they
want to model something that can be expressed as Event/Nonevent or something
that has two possible es, such as the examples given er et al. (1996) as
financial status of a firm (sound status, headed toward insolvency) or blood pressure
status (high blood pressure, not high blood pressure).
The logistic transformation in this manual provides the basis for the linear logistic
model. It is ideal in that it restricts predictions to the interval (0,1), which is
suitable for proportions and can be easily transformed into a linear form. The
function takes on a monotonically increasing or decreasing S shape that frequently
occurs in Dose/Response and other survival rate studies. It is an increasingly
popular alternative to discriminant analysis because it relaxes some of the
assumptions.
Logistic Regression in
STATGRAPHICS Plus
STATGRAPHICS Plus allows you to enter dichotomous response data in one of two
forms:
· the dependent variable can consist of individual observations of either a 0 or a 1,
representing the actual e for each individual case
· the dependent variable can contain probabilities between 0 and 1 that are the
proportions of esses for groups of observations whose sizes can vary.
The independent or explanatory variables can be either categorical or continuous.
The program creates indicator variables for categorical factors that can simplify the
use of the model when making predictions for future cases. There are several
statistics and graphs that help to assess the accuracy and usefulness of the model.
To access the analysis, choose SPECIAL... ADVANCED REGRESSION... LOGISTIC REGRESSION...
from the Menu bar to displ