Prediction Interval vs Confidence Interval

Both intervals measure uncertainty, but answer different questions:

  • Confidence Interval (CI): Uncertainty about the mean prediction — how well does the model estimate the average?
  • Prediction Interval (PI): Uncertainty about a single new observation — always wider than CI because it adds σ²
Formulas:
  • CI = ŷ ± t · √(Xₙ(XᵀX)⁻¹Xₙᵀ · σ²)
  • PI = ŷ ± t · √(Xₙ(XᵀX)⁻¹Xₙᵀ · σ² + σ²)
Rule: Wide PI → data is noisy. Narrow CI → model mean is stable.
Try it in the app
Try: Make a prediction near the edge of your data range → observe the wide PI vs the tighter CI shown in the prediction results.
Watch this on our YouTube channel

Run this on your own data

Free tier, no signup needed. Works offline in the browser.

Open the app