Mean

Statistic accumulating the per-column mean, ignoring NaN values. It is a base statistic: the normalizer reads it as the centering offset, and other quantities such as the coefficient of variation build on it. Its parallel merge relies on the count statistics.

Selected as mean. Takes no configuration.