plot_influence()

Diagnostic scatter plots for identifying influential observations in a Cox model.

Usage

Source

plot_influence(
    cox,
    *,
    highlight=3,
    panels=None,
    title=None,
    backend="altair",
    width=240,
    height=200,
)

Produces a horizontal row of panels plotting key diagnostics against the linear predictor. The most influential observations (by likelihood displacement) are highlighted in red and labeled with their observation number.

Parameters

cox: Any

A fitted CoxPH model.

highlight: int = 3

Number of most influential observations to highlight and label (default 3). Set to 0 to disable highlighting.

panels: tuple[str, …] | list[str] | None = None

Which diagnostic panels to show. Each name is a column from influence_diagnostics(). The default is ("deviance", "leverage", "ld").

title: str | None = None

Optional supertitle for the combined chart.

backend: Literal["altair", "plotnine"] = "altair"

Plotting backend: "altair" (default) or "plotnine".

width: int = 240

Width of each panel in pixels (Altair) or inches (plotnine).

height: int = 200

Height of each panel in pixels (Altair) or inches (plotnine).

Returns

alt.Chart or plotnine.ggplot

A composite chart (Altair HConcatChart) or a faceted plotnine plot.

Examples

import greenwood as gw

lung = gw.load_dataset("lung", backend="polars")
y = gw.Surv.right(lung["time"], event=(lung["status"] == 2))
cox = gw.CoxPH().fit(y, lung[["age", "sex"]])

gw.plot_influence(cox)

Show only deviance residuals and leverage, highlighting the top 5:

gw.plot_influence(cox, panels=["deviance", "leverage"], highlight=5)

Use the plotnine backend for a static ggplot object:

gw.plot_influence(cox, backend="plotnine")