Skip to main content

Regression Models API

Complete reference for all 5 regression model types.

OLS (Ordinary Least Squares)​

SELECT * FROM anofox_statistics_ols(
    table_name,          -- Table with data
    target_column,       -- y (what you predict)
    predictor_columns    -- ARRAY[x1, x2, ...] (predictors)
);

Output Fields:

  • coefficient[j] - Regression coefficients
  • std_error[j] - Standard errors
  • t_statistic[j] - t-statistics
  • p_value[j] - p-values for hypothesis tests
  • r_squared - Goodness of fit (0-1)
  • adjusted_r_squared - Penalized for predictors
  • rmse - Root mean squared error
  • aic - Akaike Information Criterion
  • bic - Bayesian Information Criterion

Ridge Regression​

SELECT * FROM anofox_statistics_ridge(
    table_name,
    target_column,
    predictor_columns,
    options -- MAP_CREATE(ARRAY['lambda'], ARRAY['0.5'])
);

Options:

  • lambda - Regularization strength (0.1 to 10.0)

WLS (Weighted Least Squares)​

SELECT * FROM anofox_statistics_wls(
    table_name,
    target_column,
    predictor_columns,
    weight_column  -- Column with observation weights
);

For heteroscedastic data where variance differs.

RLS (Recursive Least Squares)​

SELECT * FROM anofox_statistics_rls(
    table_name,
    target_column,
    predictor_columns,
    options -- MAP_CREATE(ARRAY['forgetting_factor'], ARRAY['0.95'])
);

For streaming/online learning with concept drift.

Elastic Net​

SELECT * FROM anofox_statistics_elastic_net(
    table_name,
    target_column,
    predictor_columns,
    options -- MAP_CREATE(ARRAY['alpha', 'lambda'], ARRAY['0.5', '0.1'])
);

Options:

  • alpha - L1/L2 balance (0-1)
  • lambda - Regularization strength

Aggregate Versions​

All models available with _agg suffix for GROUP BY:

SELECT
    group_id,
    result.coefficient[2] as effect
FROM data
GROUP BY group_id
APPLY anofox_statistics_ols_agg(y, ARRAY[x1, x2]);

Next Steps​

🍪 Cookie Settings