pub fn adjust(p: &[f64], method: Fdr) -> Vec<f64>
Adjusted p-values (q-values) in the original order. NaN p-values stay NaN and are not counted.