Logistic Regression

Logistic Regression icon Model a binary dependent response variable as a relationship between one or more explanatory variables by fitting a logistic regression model.

Details

A binomial logistic regression models a binary dependent variable with two outcomes by one or more independent or predictor variables.

If the binary outcomes in the dependent variable are labelled '0' and '1' then fitted and predicted values of a logistic model are the log-odds for the value labelled '1'. The log-odds can be converted to a probability by the logistic function with a probability of 1 being certainly belonging to '1'.

In this module, if a dependent variable with more than two values is selected, then the values in the dependent variable will be revalued so that the first level represents a failure and the rest represent a success.

Output

The example below shows how to use the Logistic Regression module to model the diagnosis variable (response) using a number of tumour measurements (predictor variables). The diagnosis variable is binary, indicating whether the tumour is malignant or benign.

The output below shows the estimated logistic regression model for this problem. It also includes a graph of residuals plotted against the fitted values.

The module output is the printed summary of a generalized linear model glm created in R. The summary includes the estimated coefficients and their standard errors, z-score, and p-values. More details on this output can be found in the R documentation for glm.

In our example, p-values tell us that all of the tumour measurements are statistically significant for predicting the probability that a tumour is malignant.

Parameters  

Variable name Required Constraints Description
outcome_var Yes Column Input. Text, Boolean, Integer A binary variable to be used as the dependent variable in the model.
model_var1 Yes Any column other than the column chosen for outcome_var. A required first independent variable to include in the model.
model_var2 No Any column other than the column chosen for outcome_var. An optional second predictor variable.
model_var3 No Any column other than the column chosen for outcome_var. An optional third predictor variable.
model_var4 No Any column other than the column chosen for outcome_var. An optional fourth predictor variable.
model_var5 No Any column other than the column chosen for outcome_var. An optional fifth predictor variable.
include_intercept Yes Boolean Whether to include an intercept term in the model

See Also

Updated on May 5, 2021

Was this article helpful?

Related Articles

Not the solution you were looking for?
Click the link below to submit a support ticket
CONTACT SERVICE DESK