Ctrl+K

scib_metrics.kbet

Contents

scib_metrics.kbet#

scib_metrics.kbet(X, batches, alpha=0.05)[source]#

Compute kbet [Büttner et al., 2018].

This implementation is inspired by the implementation in Pegasus: https://pegasus.readthedocs.io/en/stable/index.html

A higher acceptance rate means more mixing of batches. This implementation does not exactly mirror the default original implementation, as there is currently no adapt option.

Note that this is also not equivalent to the kbet used in the original scib package, as that one computes kbet for each cell type label. To achieve this, use scib_metrics.kbet_per_label().

Parameters:

  • X (NeighborsResults) – A NeighborsResults object.

  • batches (ndarray) – Array of shape (n_cells,) representing batch values for each cell.

  • alpha (float (default: 0.05)) – Significance level for the statistical test.

Return type:

float

Returns:

acceptance_rate

Kbet acceptance rate of the sample.

stat_mean

Mean Kbet chi-square statistic over all cells.

pvalue_mean

Mean Kbet p-value over all cells.

Contents