Residuals — model comparison at a glance

When comparing cosmological models, absolute power spectra span many orders of magnitude and small differences are invisible on a log-log plot. Residuals collapse that dynamic range: a flat line means perfect agreement, and deviations stand out immediately. Every cplots plot exposes two methods — residuals() for a standalone residual panel and with_residuals() for a stacked two-panel figure with a shared x-axis.

Setup

Standalone residuals

residuals() returns a plain XYPlot — render it, overlay it, or pass it to with_residuals() yourself.

Relative residuals

(y − y_ref) / y_ref — the default and most cosmologically natural choice.

Ratio residuals

y / y_ref — equivalent to relative + 1, preferred when you want to quote fractional deviations around unity rather than zero.

Two-panel stacked figure

with_residuals() combines the main spectrum and its residuals into a single figure. The panels share an x-axis so zoom and pan stay in sync.

Selecting the reference by label instead of index makes the code self-documenting and robust to trace reordering.

External reference array

Pass a numpy array as reference to compare against a curve that is not part of the plot — for example, an analytic approximation or a result from a different code.

cplots interpolates the external reference onto each trace's x-grid (log-interpolation when the x-axis is logarithmic).