Function average_pairwise_prediction_correlation 

pub fn average_pairwise_prediction_correlation(
    per_model_predictions: &[Vec<f64>],
) -> Result<f64, EnsembleError>

Mean Pearson correlation over all pairs of model rows: the ρ̄ of AFML §6.3.1.

Pass the models’ errors (residuals), or predictions of a de-meaned target, to get the ρ̄ that bagging_ensemble_variance means; on raw predictions of a common target this mostly measures that every model tracks the target. A pair involving a constant row contributes a correlation of 0.

use openquant::ensemble_methods::average_pairwise_prediction_correlation;

// Pair correlations are +1, -1 and -1.
let rows = [vec![1.0, 2.0, 3.0], vec![2.0, 4.0, 6.0], vec![3.0, 2.0, 1.0]];
let rho = average_pairwise_prediction_correlation(&rows)?;
assert!((rho + 1.0 / 3.0).abs() < 1e-12);

Errors

• EnsembleError::TooFewModels if there are fewer than two rows.
• EnsembleError::TooFewPredictionSamples if the first row has fewer than two entries.
• EnsembleError::LengthMismatch("prediction") if the rows differ in length.