This doesn't matter for symmetric metrics like ARI and NMI but in other cases it is important that the label and cluster are passed to the right arguments. I'm not sure it's better to pass the arguments by name (which gives less chance for error by means the metric function has to use the same names) or just be clearly about which argument has to come first (and make sure it's consistent across scIB functions). Either way the metric argument documents needs updating (it still mentions group1 and group2 instead of label_key and cluster_key).
cluster_optimal_resolution()
passes arguments to the metric function by position rather than name.https://github.com/theislab/scib/blob/82444efd3e4378c487001e701a4a1209ef954aaa/scib/metrics/clustering.py#L102
This doesn't matter for symmetric metrics like ARI and NMI but in other cases it is important that the label and cluster are passed to the right arguments. I'm not sure it's better to pass the arguments by name (which gives less chance for error by means the metric function has to use the same names) or just be clearly about which argument has to come first (and make sure it's consistent across scIB functions). Either way the
metric
argument documents needs updating (it still mentionsgroup1
andgroup2
instead oflabel_key
andcluster_key
).