Handling Multicollinearity

Detect and fix correlated predictors.

Detect with VIF

SELECT
    variable,
    vif
FROM anofox_statistics_vif(
    ARRAY['marketing_spend', 'sales_team', 'ad_budget']
);

VIF > 10 = Problematic multicollinearity

Solution: Ridge Regression

SELECT
    result.coefficients[1] as intercept,
    result.coefficients[2] as effect_spend,
    result.coefficients[3] as effect_team,
    result.r_squared
FROM sales_data
APPLY anofox_statistics_ridge_agg(
    revenue,
    ARRAY[marketing_spend, sales_team],
    MAP_CREATE(
        ARRAY['lambda'],
        ARRAY['0.5']
    )
);

Lambda Selection

-- Try multiple lambda values
SELECT lambda, aic FROM (
    SELECT '0.1' as lambda UNION
    SELECT '0.5' UNION
    SELECT '1.0' UNION
    SELECT '2.0'
) lambdas,
LATERAL anofox_statistics_ridge_agg(...)
ORDER BY aic;

Next Steps

• Prediction Intervals — Forecasting with uncertainty
• Reference: Models — Ridge API