plot_predicted_survival()

Plot per-subject predicted survival or cumulative-hazard curves.

Usage

Source

plot_predicted_survival(
    model,
    newdata,
    *,
    type="survival",
    times=None,
    labels=None,
    xlab="Time",
    ylab=None,
    width=500,
    height=300,
    backend="altair",
)

Draws one right-continuous step curve per row of newdata, using any fitted model that exposes predict(newdata, type=..., times=..., format=...) returning a frame of per-subject curves — for example SurvivalTree, RandomSurvivalForest, ExtraSurvivalTrees, CoxPH, or CoxNet. This complements plot_survival, which draws population Kaplan-Meier curves, by visualizing how predicted risk varies across individuals.

Parameters

model: Any

A fitted estimator whose predict accepts type= and returns a time column plus one column per subject (e.g. a survival forest).

newdata: Any

Covariates for the subjects to plot (a dataframe or 2-D array), passed to model.predict.

type: str = "survival"

Curve to draw: "survival" (default) or "cumulative_hazard".

times: Any = None

Times at which to evaluate the curves. Defaults to the model’s training event times.

labels: Any = None

Optional per-subject legend labels (one per row of newdata).

xlab: str = "Time"

X-axis label (default "Time").

ylab: str | None = None

Y-axis label. Defaults to "Survival probability" or "Cumulative hazard" by type.

width: int = 500

Plot dimensions in pixels (defaults 500x300).

height: int = 500

Plot dimensions in pixels (defaults 500x300).

backend: str = "altair"

Plotting backend. Currently only "altair" is supported.

Returns

altair.Chart

An interactive Altair chart with one colored step curve per subject.

Examples

Fit a random survival forest and plot survival curves for a few subjects:

import greenwood as gw

# Load data and build a right-censored response
lung = gw.load_dataset("lung", backend="pandas").dropna(subset=["ph.ecog", "ph.karno"])
y = gw.Surv.right(lung["time"], event=(lung["status"] == 2))
cols = ["age", "sex", "ph.ecog", "ph.karno", "wt.loss"]

# Fit a forest and plot per-subject predicted survival
rsf = gw.RandomSurvivalForest(n_estimators=100, random_state=0).fit(y, lung[cols])
gw.plot_predicted_survival(rsf, lung[cols][:4])