scib.metrics.ari

scib.metrics.ari(adata, group1, group2, implementation=None)

Adjusted Rand Index

The function is symmetric, so group1 and group2 can be switched For single cell integration evaluation the comparison is between predicted cluster assignments and the ground truth (e.g. cell type)

Parameters

• adata – anndata object

• group1 – string of column in adata.obs containing labels

• group2 – string of column in adata.obs containing labels

• implementation – if set to ‘sklearn’, uses sklearn’s implementation, otherwise native implementation is taken