yasa.SleepStatsAgreement.variance_components#

property SleepStatsAgreement.variance_components[source]#

Between- and within-subject components of the variability of the differences.

A pandas.DataFrame with one row per sleep statistic and the following columns:

  • sd_between — SD of the mean difference of each subject, i.e. the stable offsets of the observed scorer for a given subject. NaN when no subject has several valid nights for that statistic (e.g. without subjects).

  • sd_within — SD of the differences across the nights of a subject. NaN as above.

  • sd_total — sqrt(sd_between² + sd_within²), the SD of a new night of a new subject, which defines the limits of agreement (bias ± 1.96 × sd_total). Without repeated measures, this is the sample SD of the differences.

  • n_subjects and n_sessions — number of subjects and sessions with a valid value.

The components are estimated with the one-way random-effects ANOVA of Bland & Altman (2007), which handles unequal numbers of nights per subject. Together, they tell whether the disagreement between scorers comes mostly from stable subject-level offsets or from night-to-night variability (Parker et al., 2016).

Added in version 0.8.1.