jsurvival

Survival analysis: Kaplan–Meier, Cox, competing risks, multi-state, cure models, and ML survival

← Home · Onboarding · jamovi overview

What it does

The survival-analysis arm of the ClinicoPathJamoviModule. Handles almost every time-to-event question that comes up in clinicopathological research.

Core methods

  • Kaplan–Meier and Nelson–Aalen estimators.
  • Turnbull NPMLE for interval-censored data.
  • Log-rank and stratified log-rank tests.

Modeling

  • Cox proportional hazards (with coxph diagnostics baked in).
  • Time-varying covariates and time-varying effects.
  • Accelerated failure time (AFT) models.
  • Frailty / shared-frailty models.
  • Landmark analysis.

Specialized applications

  • Competing risks (Fine–Gray, cause-specific).
  • Multi-state models (illness–death and beyond).
  • Recurrent event analysis.
  • Cure models (mixture and non-mixture).

Oncology-specific

  • Progression-free survival with RECIST-aware handling.
  • IPCW for treatment switching.
  • Tumor growth kinetics.

Machine learning

  • Survival trees and survival random forests.
  • Gradient-boosted survival (XGBoost Cox, GBM-Cox).
  • Regularized regression (LASSO / Ridge / Elastic Net) for high-dimensional covariates.

When to use

  • Any paper with a time-to-event outcome.
  • When you need publication-ready KM curves with risk tables and p-value annotations.
  • When competing events or interval censoring are present (use these; don’t pretend they aren’t).

Repos

Quick start in jamovi

  1. Open the cohort dataset.
  2. Analyses → ClinicoPath → Survival → Kaplan–Meier.
  3. Select time variable, event variable (with event level), and grouping variable.
  4. Publication-ready plot + risk table appear on the right.

Quick start in R

library(jsurvival)
km <- survival_function(
  data = mydata,
  time = "os_months",
  event = "os_event",
  group = "treatment"
)

Pitfalls

  • Event coding. The event variable must be a factor with the event level explicitly specified. Silent 0/1 coding errors are the single most common mistake in the group.
  • Proportional hazards. Don’t report Cox HRs without running the Schoenfeld test. The module prints it by default; don’t hide it.
  • Competing events matter. If a non-trivial fraction of patients die of other causes, cause-specific or Fine–Gray is the correct model — not naïve KM on disease-specific death.