numeraire_graphics.plot_ic_decay

numeraire_graphics.plot_ic_decay(results: DataFrame, *, horizon: str = 'horizon', metric: str = 'ic', smooth: bool = False) → plotnine.ggplot

The information coefficient plotted against a caller-assembled forecast-horizon axis.

Consumes the ic rows ICEvaluator emits (metric="ic" by default; "ic_ir" or "ic_t" read the same way). A single ForecastOutput carries one horizon, so its ic row is scalar; the decay curve is assembled by the caller running forecasts at several horizons and tagging each ic row with a numeric horizon column, then stacking them. The result schema has no horizon column — exactly the caller-supplied-axis pattern of plot_complexity_curve() — so horizon is an explicit argument and its absence raises.

Rows are sorted along horizon and drawn as a geom_line + geom_point coloured by method, over a zero reference line (an IC decaying toward zero as the horizon lengthens is the figure’s whole point). smooth=True overlays a light linear trend (geom_smooth) per method for the eye.