Simulation laboratory

Change the assumptions. Watch what responds.

Two small experiments that run entirely in your browser. Start with the highlighted control, change one thing at a time, then decide what the evidence supports.

01 · Simulate a trial02 · Compare synthesizers

Experiment 01 · Knowledge-driven simulation

Simulate a two-arm survival trial

No patient data needed: every patient is generated from the assumptions on the left. This is the method used for sample-size and design decisions.

AssumeSimulateInterpretDecide

Explore the evidence

Kaplan–Meier curves

PlaceboActive
–
Estimated hazard ratio
–
Events observed
Median (months):
–
Power across repeated trials
Press “Estimate power”
Interpret

Experiment 02 · Fidelity versus privacy

Compare three ways of making synthetic patients

The rings are 250 real patients (a teaching dataset). The filled dots are 250 synthetic patients made by the method you choose. The scores show fidelity and privacy pulling against each other.

ChooseGenerateCompareDecide

Explore the evidence

Real and synthetic patients

RealSynthetic
–
Correlation, synthetic vs real
–
Difference in distributions (KS, lower is better)
–
Near-copies: closer to a real patient than real patients are to each other (lower is safer)
Interpret