Account for noisy predictors, imperfect outcomes, misclassification, and uncertain instruments. Use calibration data and sensitivity analysis to show how measurement error affects estimates.
