Small interactive apps demonstrating specific topics from the course. Each link opens a standalone page.
Click to add points and watch the OLS line and residual sum of squares update live.
See the sum of squared residuals as a genuine function of β₀ and β₁ — a rotatable 3D bowl and a matching contour map, with sliders marking exactly where your chosen coefficients sit on it.
1000 simulated samples confirm the theoretical sampling distributions of β̂₀ and β̂₁, and a forest plot of confidence intervals shows coverage matching the nominal confidence level.
Residuals vs. fitted, Normal Q-Q, a histogram of standardized residuals, and residuals vs. leverage — all on the same simulated dataset, with draggable points and severity sliders for heteroscedasticity, nonlinearity, and skew.
Drag to rotate a 3D scatter of two predictors and an outcome, with the OLS-fitted regression plane rendered as a wireframe mesh.
Two groups, a no-interaction model (forced parallel lines) and an interaction model (separate slopes) fit side by side — turn up the true slope difference and watch the parallel-lines model visibly break.
Slide the correlation between two predictors and watch 150 resampled coefficient estimates stretch into an unstable ridge.
Draggable two-class points with a live decision boundary and probability contours, linked to a sigmoid curve panel showing each point's fitted probability.
Thirty multiple-choice/true-false questions reviewing matrix notation, the hat matrix, the F-test family, adjusted R², feature engineering, multicollinearity, and leverage — click an answer to see if you're right, and why.
Drag a candidate mean across a sample from a known-variance Normal and watch its density curve, the likelihood, and the log-likelihood all respond together — both curves peak at exactly the sample mean.
Watch Gradient Descent, Momentum, AdaGrad, Adam, and Newton-Raphson roll a ball downhill on a rotatable 3D surface and its contour map — pick a convex bowl, a multi-minima surface, or a saddle point, and compare how each method behaves.