Every compared recording is analysed with the same shared parameters, so the panels are like-for-like. Dated settings — analysis window, exclusions, τ-fit ranges, shift day and hand-corrected markers — belong to the active recording, so by default each compared animal is analysed over its own full record (change this under “Compare recordings” in the sidebar).

Forking is non-destructive: A holds the parameters you forked from, B is the variant. The fork marked Active analysis is the one the sidebar controls edit — the other is parked with the result it was last analysed with, so changing a parameter can never overwrite it. Promote adopts a fork as the analysis and leaves A/B mode; Discard returns to the settings you forked from.

Assign each recording a group and (optionally) a subject id on the Compare tab. Ungrouped recordings stay in Compare but are excluded here. Recordings sharing a subject id aggregate before testing, so the effective n counts subjects, not recordings — the field's most common statistical error is treating slices or cells as independent replicates.

Parameter-level comparison (period, amplitude, acrophase) is often a stronger claim than a pointwise bar (Hughes et al. 2017). Fit each group separately and compare the fitted parameters when the biological question is a shift in one of them, not a bin-wise difference in the raw trace.