scipp.nanmean#

scipp.nanmean(x, dim=None, *, out=None)#

Element-wise mean over the specified dimension ignoring NaNs.

If the input has variances, the variances stored in the ouput are based on the “standard deviation of the mean”, i.e., \(\sigma_{mean} = \sigma / \sqrt{N}\). \(N\) is the length of the input dimension. \(\sigma\) is estimated as the average of the standard deviations of the input elements along that dimension.

See scipp.sum() on how rounding errors for float32 inputs are handled.

Parameters
Raises

If the dimension does not exist, or the dtype cannot be summed, e.g., if it is a string.

Returns

Union[Variable, DataArray, Dataset] – The mean of the input values.

Seealso

scipp.mean().