What is Binary Logistic Regression?

Binary Logistic Regression is a statistical method used to predict the probability of a binary outcome (an event that has only two possible states, such as Yes/No, Pass/Fail, or 0/1). Unlike Linear Regression, which fits a straight line, Logistic Regression fits an "S-shaped" logistic curve (sigmoid) to the data, ensuring that predicted probabilities always fall between 0 and 1.

This tool uses the Logit transformation to model the relationship between one or more independent variables (predictors) and the probability of your target event occurring.

When to use it?

You should use Binary Logistic Regression instead of Linear Regression when:

• Binary Outcome: Your dependent variable is categorical with two levels (e.g., Churn vs. Retain, Disease vs. Healthy).
• Probability Estimation: You need to know the likelihood of an event, not just the classification.
• Odds Ratios: You want to understand how a predictor impacts the likelihood of an outcome (e.g., 'Does increasing price by $1 multiply or decrease the odds of a sale?').
• Medical & Social Science: It is the standard for determining risk factors in clinical studies.

How our tool works (Accuracy & Stack)

Our tool is powered by a robust Python backend (running on Google Cloud Functions and Pyodide) utilizing the statsmodels.Logit library. We calculate Maximum Likelihood Estimates (MLE) to derive coefficients, ensuring professional-grade accuracy comparable to R, SPSS, or SAS.

We go beyond basic coefficients by including advanced diagnostics usually missing from online calculators:

• Variance Inflation Factor (VIF): Automatically detects multicollinearity among your predictors.
• Cook’s Distance: Identifies influential data points (outliers) that might be skewing your model.
• Pseudo R-Squared: Reports McFadden’s R², AIC, and BIC for model fit comparison.

Visualizations & Model Evaluation

To help you interpret the results, we generate interactive Plotly charts:

• ROC Curve: Visualizes the trade-off between Sensitivity and Specificity. Includes the AUC (Area Under the Curve) score.
• Confusion Matrix Heatmap: A dynamic view of True Positives, False Positives, and overall accuracy. Includes derived metrics like Precision, Recall (Sensitivity), and F1-Score.
• S-Curve Plot: Visualizes the predicted probabilities against the sorted data to show the logistic fit.

Input Requirements & Limitations

To ensure speed and responsiveness for the interactive diagnostics:

• Row Limit: Supports up to 5,000 rows.
• Column Limit: Supports up to 30 predictor variables (columns).
• Data Format: The Dependent Variable (Y) must be strictly binary (0 and 1).
• Optional: You can include a Weight column for weighted analysis.

How to Use the Binary Logistic Regression Calculator

Our tool is designed to be intuitive, even if you are new to statistical modeling. Follow these steps to perform your analysis:

1. Prepare Your Data: Ensure your Dependent Variable (Y) is strictly binary.
   • Use 1 for the "Event" (e.g., Success, Yes, Churn, Disease).
   • Use 0 for the "Non-Event" (e.g., Failure, No, Retain, Healthy).

   Predictor variables (X) can be continuous (e.g., Age, Price) or categorical (coded as numbers).

2. Import Data: You can either copy-paste your data directly into the grid or upload a CSV file. The tool supports up to 5,000 rows.
3. Select Variables:
   • Y Variable: Select the column representing your binary outcome.
   • X Variables: Check the boxes for one or more independent variables you want to use for prediction.
4. Configure Settings (Optional):
   • Cutoff Value: (Default 0.5) The probability threshold used to classify a prediction as "1".
   • Significance Level (α): (Default 0.05) Adjust for 90%, 95%, or 99% confidence intervals.
5. Analyze & Interpret: Click "Calculate Results." Use the Summary Tab for p-values and coefficients, and the Confusion Matrix tab to assess prediction accuracy.

Differences & Similarities with Other Statistical Tools

Choosing the right statistical test is critical. Here is how Binary Logistic Regression compares to other common methods available on LearnBin Lab.

1. Binary Logistic Regression vs. Multiple Linear Regression

These are the two most common forms of regression, but they serve different purposes.

• Similarity: Both model the relationship between Independent Variables (X) and a Dependent Variable (Y). Both provide coefficients ( β) and P-values to test statistical significance.
• Difference (Outcome): Linear Regression predicts a continuous number (e.g., "Sales will be $50,000"). Logistic Regression predicts the probability of a category (e.g., "75% chance of Sale").
• Difference (Shape): Linear Regression fits a straight line. Logistic Regression fits an S-shaped (Sigmoid) curve to keep predictions between 0 and 1.

2. Binary Logistic Regression vs. Chi-Square Test

Both tools handle categorical data, but they answer different questions.

• Similarity: Both can analyze the relationship between two categorical variables (e.g., Gender vs. Voting Preference).
• Difference: The Chi-Square Test only tells you if an association exists (Are they related?). Logistic Regression allows you to predict the outcome and control for other variables simultaneously (e.g., "How does Gender affect Voting, holding Age constant?").

3. Binary Logistic Regression vs. ANOVA

• Similarity: Both are used to compare groups.
• Difference: ANOVA tests if the Means of a continuous variable differ across groups. Logistic Regression tests if the Probability of a binary event differs based on the inputs.