pub fn ewm(values: &[f64], alpha: f64) -> Vec<f64>Expand description
Exponentially weighted mean with smoothing factor alpha ∈ (0, 1].
ewm[i] = alpha * values[i] + (1 - alpha) * ewm[i-1], seeded with the
first non-NULL value — pandas’ ewm(alpha=…, adjust=False).mean().
A NULL contributes no observation: the state carries across it unchanged
and is what the row reports, exactly as rolling_mean still reports its
window’s mean at a NULL row. The statistic is a property of the samples
seen so far, and a row with no sample does not undefine it. Folding the
NaN into the recurrence, which is what this used to do, made every row
after the first NULL NULL as well.
Matches pandas’ ewm(alpha=…, adjust=False, ignore_na=True).mean().