A computer program for linear logistic regression analysis

✍ Scribed by Elisa T Lee

• Publisher: Elsevier Science
• Year: 1974
• Weight: 545 KB
• Volume: 4
• Category: Article
• ISSN: 0010-468X
• DOI: 10.1016/0010-468x(74)90011-7

✦ Synopsis

Given a set of binary data (success of failure) and measurements of p explanatory variables corresponding to each case, a computer program has been written to perform regression analysis utilizing Cox's linear logistic model. A prototype example is to assess any dependence of the probability of response or complete remission on patient characteristics such as age, temperature, and white blood count etc. The program has two versions and they are called LOGIST and STLOG. LOGIST fits all the p explanatory variables at the same time while STLOG follows a forward selection procedure so that the first variable determined is the most important single characteristic in predicting probability of response, the second variable is second most important after the first one has been included in the equation, and so on. The program gives (1) maximum likelihood estimates and interval estimates for the regression coefficients, (2) significance level for hypothesis tested, and (3) predicted probability of success and residual for each case.

Regression methods

Linear logistic models Binary data