Survival Curve Builder

Interactive Kaplan-Meier curves, log-rank tests, and Cox model fits, all from a single data input panel

Purpose

Survival data has a distinctive shape: time, event indicator, covariates. The Survival Curve Builder lets a reader upload such data (or simulate it) and explore Kaplan-Meier estimation, the log-rank test for group comparisons, and Cox proportional hazards regression, with each output updating as inclusion criteria and covariates change.

User inputs

• Dataset (built-in lung, ovarian, pbc, or user-uploaded CSV)
• Time and event columns
• Grouping variable for curve stratification
• Covariates for the Cox model with continuous/categorical toggles
• Subset filters with live row count

Outputs

• Kaplan-Meier curves with pointwise 95% bands and a risk table below
• Log-rank test result and, if requested, weighted log-rank variants
• Cox regression table with hazard ratios, CIs, p-values, and overall model tests
• Schoenfeld-residual diagnostic plot for each covariate, with a proportional-hazards p-value
• Predicted survival curves at user-specified covariate profiles

Didactic value

Dragging the stratification variable on and off turns a single pooled curve into a comparison, making the log-rank test’s null hypothesis concrete. Watching hazard ratios move as covariates are added teaches confounding adjustment in a way a static example cannot.

Embedded in

• survival-analysis/kaplan-meier.md
• survival-analysis/log-rank-test.md
• survival-analysis/cox-regression.md

Source code

Local: apps/11-survival-curve-builder/

Run with:

shiny::runApp("apps/11-survival-curve-builder")