A bivariate measurement error model for semicontinuous and continuous variables: Application to nutritional...

Area of Science:

• Biostatistics
• Nutritional Epidemiology
• Measurement Error Correction

Background:

• Semicontinuous data with many zeros are common in biostatistics, particularly in nutritional epidemiology.
• Accurate assessment of dietary intake is crucial for understanding disease-risk relationships.
• Food frequency questionnaires introduce substantial measurement error in dietary assessment.

Purpose of the Study:

• To develop a bivariate model for simultaneously analyzing energy-adjusted dietary components and energy intake.
• To apply this bivariate model within the regression calibration framework for measurement error correction.
• To improve the estimation of risk models in nutritional epidemiology.

Main Methods:

• Developed a novel bivariate model for semicontinuous and continuous variables.
• Applied the bivariate model to the regression calibration approach for correcting dietary measurement error.
• Utilized short-term reference measurements for estimating regression calibration predictors.
• Illustrated the methodology with data from the NIH-AARP Diet and Health Study.

Main Results:

• The proposed bivariate model enables simultaneous modeling of dietary components and energy intake.
• The methodology effectively addresses measurement error in energy-adjusted dietary components.
• The approach enhances the accuracy of risk prediction models in nutritional studies.

Conclusions:

• The developed bivariate model is a valuable tool for analyzing semicontinuous dietary data in nutritional epidemiology.
• This method improves the reliability of risk estimates by correcting for measurement error.
• The approach has significant implications for public health research linking diet and disease.