Linear regression estimates an intercept and predictor coefficients by minimising squared residuals. Logistic regression estimates event probabilities through the logit link.
Linear model: ŷ = β₀ + β₁x₁ + β₂x₂ + ··· + βₚxₚ Logistic model: P(Y = 1 | X) = 1 / (1 + e⁻ᶻ), where z = β₀ + β₁x₁ + ··· + βₚxₚ AIC = 2k − 2ln(L) BIC = k ln(n) − 2ln(L) Adjusted R² = 1 − (1 − R²)(n − 1)/(n − k − 1)
Paste a dataset with headers, then inspect its variables. Choose the outcome, predictors, selection method, and model settings.
Run the calculator and review the final equation. Check diagnostics, validation results, influence measures, and warnings before interpreting selected predictors.
Use prediction inputs for new observations after fitting. Export tables or copy the complete statistical report for further work.
| Sales | Advertising | Price | Stores | Region |
|---|---|---|---|---|
| 132 | 18 | 9.5 | 12 | North |
| 145 | 22 | 9.1 | 14 | South |
| 151 | 25 | 8.8 | 15 | North |
| 160 | 28 | 8.6 | 17 | West |
What is stepwise regression?
It adds or removes predictors using a chosen statistical criterion. The procedure searches many candidate models automatically.
Which selection method should I choose?
Forward selection begins small, while backward elimination begins with all predictors. Bidirectional selection can add and remove terms.
Can categorical predictors be included?
Yes. Select categorical variables and the calculator creates dummy variables using your reference-category rule.
Why can stepwise results be unstable?
Small data changes can alter which predictors enter. Correlated predictors and limited samples increase instability.
What do AIC and BIC measure?
Both balance model fit against complexity. Lower values usually indicate a preferable candidate model.
What does a high VIF mean?
It suggests a predictor overlaps strongly with other predictors. Coefficients may become unstable or difficult to interpret.
Should influential observations be removed?
Not automatically. Investigate data quality and compare sensitivity before excluding any observation.
Why should I validate the model?
Validation estimates performance on unseen data. It helps reveal overfitting caused by repeated model selection.
Can this calculator prove causation?
No. Regression describes conditional associations unless supported by a suitable design and defensible causal assumptions.