A nonparametric approach to modeling birth weight in the presence of gestational age error

Regression problems in which predictors are measured with error have been studied by statisticians and researchers for many years. Measurement error in predictors creates biases in estimated regression coefficients, and hence models that address this are extremely important. These models are especially important in perinatal research since errors in gestational age can have serious effects.;The presence of measurement error in gestational age can lead to poor estimation of fetal growth and risk of mortality and morbidity, and can compromise statistical analyses [32, 39]. Since various obstetric choices are made based on birth weight distributions by gestational age, it is important to obtain plausible birthweight-gestational-age combinations.;Berry et al. [3] propose a Bayesian approach to modeling a flexible regression function in the presence of measurement error, where the regression function is modeled using smoothing splines and regression P-splines. These methods are applied to population-based data from U.S. birth certificates, which results in realistic birthweight-gestational age combinations.