Regression calibration method for correcting measurement-error bias in nutritional epidemiology.

摘要: Regression calibration is a statistical method for adjusting point and interval estimates of effect obtained from regression models commonly used in epidemiology bias due to measurement error assessing nutrients or other variables. Previous work developed use estimating odds ratios logistic regression. We extend this here incidence rate Cox proportional hazards slopes linear-regression models. appropriate when gold standard available validation study linear with constant variance applies replicate measurements are reliability random within-person can be assumed. In paper, the illustrated by correction describing relations between breast cancer dietary intakes vitamin A, alcohol, total energy Nurses' Health Study. An example using based on estimation relation ultradistal radius bone density caffeine, calcium, Massachusetts Women's Software implementing these methods uses SAS macros.