The following code raises an error

import dask.array as da
import sparse
import numpy as np
x = da.from_array(sparse.random((10, 10), nnz=20, data_rvs=np.random.random))
x.std(axis=0).compute()

The current issue is that dask.array.reductions.moment_agg has a line denominator[denominator < 0] = np.nan, but in this case denominator is a sp.COO matrix which does not allow item assignment. A possible solution is to use denominator = np.where(denominator<0, np.nan, denominator) instead of the direct assignment .