Prediction Intervals

Forecast with uncertainty quantification.

Generate Predictions

SELECT
    scenario.spend,
    pred.point_estimate as predicted_revenue,
    pred.lower_95,
    pred.upper_95
FROM (
    SELECT ARRAY[10000, 25000, 50000] as spending
) scenario,
LATERAL anofox_statistics_ols_predict_interval(
    'historical_data',
    'revenue',
    ARRAY['marketing_spend'],
    STRUCT_PACK(marketing_spend := scenario.spending[idx]),
    0.95
) as pred;

Confidence vs. Prediction

Mean forecast: $100K ± $10K (confidence interval)
Individual forecast: $100K ± $30K (prediction interval)

Use Cases

• 95% intervals for standard reporting
• 80% intervals for planning (less uncertainty)
• 99% intervals for risk management

Next Steps

• Production Deployment — Monitoring
• Reference: Inference — Prediction API