Interpreting confidence and prediction intervals

Advanced Probability: Uncertainty in Data

Maarten Van den Broeck

Senior Content Developer at DataCamp

Why do we need intervals?

• Data-driven decisions involve uncertainty
• Variability in the data and outcomes
• Confidence and prediction intervals help communicate uncertainty

Confidence intervals

• Reliability of estimated statistical parameter
• Based on sampling process

Confidence interval example

Prediction intervals

• Range for individual new data point
• Based on statistical model
• Typically wider

Prediction interval example

Confidence vs. prediction intervals

Confidence interval

• Range for a statistical parameter
• Based on sampling process
• Typically narrower

Prediction interval

• Range for an individual observation
• Based on a predictive model
• Typically wider

Example: risk management

• Return on investment portfolio
• Average annual return of 8%

Example: risk management

Interval Type	Purpose	95% Interval Range
Confidence Interval	Portfolio's average return estimate	7% to 9%
Prediction Interval	Single investment return estimate	4% to 12%

Let's practice!

Advanced Probability: Uncertainty in Data