curerate 0.2.0

Cure rate models for time-lagged conversion analysis

curerate

curerate is a Python library for fitting cure rate models — statistical models for time-lagged conversions where some fraction of the population will never convert.

It is a modernised fork of convoys by Erik Bernhardsson.

Installation

pip install curerate

Quickstart

import curerate as cr

# Your DataFrame has timestamp columns — curerate handles the rest
model = cr.ConversionModel(dist="weibull")
model.fit(
    df,
    created="signed_up_at",
    converted="converted_at",
    group="plan",           # optional
)

# Predict conversion rates at specific time points
model.predict(t=[7, 14, 30, 60, 90])

# What fraction will ever convert?
model.predict_final()

# Summary table
model.summary()

# Plot conversion curves
model.plot(t_max=90)

API

ConversionModel(dist, mcmc, hierarchical)

Parameter	Default	Description
dist	"weibull"	Distribution: "exponential", "weibull", "gamma", "generalized_gamma", or "kaplan_meier"
mcmc	False	Enable MCMC sampling for confidence intervals
hierarchical	True	Regularise parameters across groups

model.fit(df, *, created, converted, group=None, now=None, t_unit="days")

Fits the model to a DataFrame. Rows where converted is NaT/NaN are treated as censored (not yet converted).

model.predict(t, group=None, ci=None) → DataFrame

Returns predicted conversion rates at times t. Columns: t, group, conversion, and optionally conversion_low/conversion_high when ci is set and mcmc=True.

model.predict_final(ci=None) → DataFrame

Returns the asymptotic conversion rate (the fraction that will ever convert). Columns: group, conversion, and optionally conversion_low/conversion_high.

model.summary()

Prints a summary table with per-group statistics.

model.plot(t_max=90, ci=None, ax=None) → Axes

Plots conversion curves for all groups. Returns a matplotlib Axes.

License

MIT