Probit/Logit Regression - Input File

There are two ways to set up a data file for a probit or logit regression analysis. First, a data file can contain raw data, with a coding variable as the dependent variable (containing the two codes that indicate the two levels of the dichotomous dependent variable). If the data are arranged in this manner, then select Codes and no counts in the Input file contains box on the Logistic Regression (Logit) and/or Probit Regression data file - Quick tab.

Alternatively, you can enter frequencies directly into the file. This way of arranging the data file is often more efficient when the data contain failure (or survival) rates as a function of time. For example, Cox (1970) reported the following data on tests of objects for failures after certain times. Variable Failure is an indicator variable (0=no failure, 1=failure), and variable Count represents the respective number of objects that did or did not fail in the respective time interval:

Time

Failure

Count

7

0

55

14

0

155

14

1

2

27

0

152

27

1

7

51

0

13

51

1

3

Thus, at time 7 there were no failures among 55 objects. At time 14, there were 2 failures (coded 1) and 155 non-failures (coded 0, for a total of 157 objects). At time 27 there were 7 failures and 152 non-failures, etc. To analyze this data set, select Codes and counts from the Input file contains box on the Logistic Regression (Logit) and/or Probit Regression data file - Quick tab, specify Time as the independent variable (in the first list of the Specify Variables dialog), Failure as the dependent variable (in the second list), and variable Count as the variable containing the counts (in the third list). Note that the total frequencies reported in the Results dialog box will be the actual frequencies, that is, taking into account the counting variable.