NaN in single component drops the whole sample even if other components are OK. This is too restrictive in many cases.
Avarages could be computed in Numpy ommiting the NaNs nanmean, that is compute mean for individual components of the vector quantity.
Consider introducing a function that is true if the sample has any Nan, then ti could be used with select mechanism to drop samples containing any Nan.
NaN in single component drops the whole sample even if other components are OK. This is too restrictive in many cases.
Avarages could be computed in Numpy ommiting the NaNs
nanmean, that is compute mean for individual components of the vector quantity.Consider introducing a function that is true if the sample has any Nan, then ti could be used with select mechanism to drop samples containing any Nan.