Statistical technique that repeatedly samples from observed data to estimate uncertainty without making strong assumptions.