Quantifying uncertainty in epidemiological models