Although our estimator is derived for an OLS estimation framework, it will be approximately correct for logistic regression35 and survival models36 as long as the coefficient on the standardized additive SNP factor, βg, is not too large. For example, applying a measurement-error correction that would be correct for OLS will be a very accurate approximation for the coefficient in a survival model when the hazard ratio associated with a one-standard deviation difference in the variable measured without error is 1.1136. However, the correction is roughly 20% too small when the hazard ratio is 1.6536.
