Interpretation of 'score' output in sc.tl.rank_genes_groups()

Hi there,

Could someone provide an explanation of the meaning of the ‘score’ field in the output of sc.tl.rank_genes_groups()? It seems broadly correlated with both p-value and logfoldchange, but I can’t understand the connection it has.

The documentation page says “Structured array to be indexed by group id storing the z-score underlying the computation of a p-value for each gene for each group. Ordered according to scores.” but I don’t understand what this means.

I was thinking that perhaps it would be a score from a statistical test, but in this case I would expect the rank of the abs(scores) to be inverse to the rank of the p-values - however I find that there are some genes which have higher scores but higher p-values, indicating that this might not be the case.

Clarification on this would be much appreciated, thanks in advance.