Function calculate_bin_stats

src/memory/processing/hvg/mod.rs:293–335

Calculate mean and standard deviation of dispersions within each expression bin. For the Seurat method, this establishes the expected mean-variance relationship by computing statistics within bins of similar expression levels. ## Parameters `log_dispersions` - Log-transformed dispersion values (variance/mean) `bin_indices` - Bin assignment for each gene `n_bins` - Total number of bins ## Return

(
    log_dispersions: &[f64],
    bin_indices: &[usize],
    n_bins: usize,
)

/// - Single-gene bins: Mean = gene value, std = NaN (handled by postprocessing)
/// - Uses sample standard deviation (n-1 denominator)
fn calculate_bin_stats(
    log_dispersions: &[f64],
    bin_indices: &[usize],
    n_bins: usize,
) -> anyhow::Result<(Vec<f64>, Vec<f64>)> {
    let mut bin_values: Vec<Vec<f64>> = vec![Vec::new(); n_bins];

    // Collect values for each bin (excluding NaN)
    for (i, &bin_idx) in bin_indices.iter().enumerate() {
        let disp = log_dispersions[i];
        if !disp.is_nan() && bin_idx < n_bins {
            bin_values[bin_idx].push(disp);
        }
    }

    let mut bin_means = vec![0.0; n_bins];
    let mut bin_stds = vec![0.0; n_bins];

    for bin_idx in 0..n_bins {
        let values = &bin_values[bin_idx];

        if values.is_empty() {
            bin_means[bin_idx] = f64::NAN;
            bin_stds[bin_idx] = f64::NAN;
        } else if values.len() == 1 {
            // Single gene in bin - Python sets std to NaN
            bin_means[bin_idx] = values[0];
            bin_stds[bin_idx] = f64::NAN;
        } else {
            // Calculate mean
            let mean = values.iter().sum::<f64>() / values.len() as f64;
            bin_means[bin_idx] = mean;

            // Calculate standard deviation
            let variance =
                values.iter().map(|&x| (x - mean).powi(2)).sum::<f64>() / (values.len() - 1) as f64;

            bin_stds[bin_idx] = variance.sqrt();
        }
    }

    Ok((bin_means, bin_stds))
}

/// Normalize dispersions by subtracting expected values and dividing by standard deviation.
///