Regression

Build predictive models

Regression

Regression models predict outcomes based on input variables. StatKit supports linear and logistic regression.

Types of Regression

Linear Regression

Predict continuous outcomes.

Example: Predict house prices based on size, location, age.

Logistic Regression

Predict binary outcomes.

Example: Predict whether a customer will buy (yes/no) based on demographics.

Data Requirements

Linear Regression

  • Dependent Variable: Continuous numeric outcome
  • Independent Variables: One or more numeric predictors
  • Sample Size: At least n+1 observations (where n = number of predictors)

Logistic Regression

  • Dependent Variable: Binary or categorical with exactly 2 values
  • Independent Variables: One or more numeric or categorical predictors
  • Sample Size: At least n+1 observations per category

How to Run

  1. Select “Regression” from the test dropdown
  2. Choose your CSV file
  3. Select your dependent variable
  4. Choose independent variables
  5. Click “Run Test”

Interpreting Results

Linear Regression

  • R-squared: Proportion of variance explained (0 to 1)
  • RMSE: Root Mean Square Error - prediction accuracy
  • Coefficients: Change in outcome per unit change in predictor
  • P-values: Statistical significance of predictors

Logistic Regression

  • Pseudo R-squared: Measure of model fit
  • Coefficients: Change in log-odds per unit change in predictor
  • P-values: Statistical significance of predictors

Model Fit

  • Linear: Higher R-squared indicates better fit
  • Logistic: Higher pseudo R-squared indicates better fit
  • P-values: < 0.05 indicates significant predictors

Assumptions

Linear Regression

  • Linearity: Relationship between variables is linear
  • Independence: Observations are independent
  • Homoscedasticity: Constant variance of residuals
  • Normality: Residuals are normally distributed

Logistic Regression

  • Independence: Observations are independent
  • No multicollinearity: Predictors are not highly correlated
  • Large sample size: At least 10 events per predictor

Export Options

  • LaTeX: Copy-paste ready tables
  • CSV: Structured data files

Need help? Check the FAQ or contact support.