Yuanyuan Luan
Yuanyuan Luan is currently a third year PhD candidate. Her dissertation focuses on developing measurement error correction methods in device-based measures of physical activity and self-reported measures of dietary intake in obesity and T2D studies under the guidance of Dr. Tekwe. As part of her research, she is working on parametric and semi parametric approaches to measurement error correction associated with high dimensional functional data. She has an MPH in Biostatistics from Texas A&M University and BA in Statistics from University of Minnesota. She can be contacted at: luany@iu.edu.
Novel Statistical Approaches for Measurement Error Correction to Evaluate the Roles of Device-based Physical Activity and Self-reported Total Caloric Intake on the Risks of Type 2 Diabetes among U.S. Adults
Yuanyuan Luan1, Roger S. Zoh1, Carmen D. Tekwe1
1Department of Epidemiology and Biostatistics, Indiana University, Bloomington, IN, USA
Although extensive work has been performed to correct for biases due to measurement errors in error-prone scalar-valued covariates in regression models, limited work has focused on addressing biases associated with functional covariates prone to errors or the combination of scalar and functional covariates prone to errors in generalized linear regression models. In this work, we develop semiparametric and parametric approaches to correct for measurement errors associated with a mixture of functional and scalar covariates prone to errors in generalized linear regression models. The methods are applied to NHANES database to assess the relationships of physical activity and total caloric intake measures with type 2 diabetes incidence among community-dwelling adults living in the United States. We treat the device-based physical activity measures as functional covariates prone to complex arbitrary heteroscedastic errors and treat total caloric intake as a scalar-valued covariate prone to error. We adjusted for age, sex, and race in our analysis. Simulation studies were conducted to establish the finite sample properties of our methods. Overall, the relationship between physical activity and the incidence of type 2 diabetes depends on the timing of physical activity. We found that the failure to account for measurement error in physical activity and total caloric intake measurements resulted in the underestimation of their relationships with type 2 diabetes incidence by 31% and 51%, respectively.
Accessibility Note: To access auto closed captions (CC) services please click on the CC icon in the bottom of the video. Manual CC are available by submitting a request for services to ATAC: (812) 856-4112 or atac@iu.edu