Indentifying Glucose-Related Dietary Patterns Accounting for Within Individual Variation

• Activity Number: 580
• Type: Contributed
• Date/Time: Wednesday, August 1, 2012 : 2:00 PM to 3:50 PM
• Sponsor: Section on Statistics in Epidemiology
• Author(s): Hsing-Yi Chang*+
• Companies: National Health Research Institutes
• Keywords: within individual variation ; dietary pattern ; diabetes ; rank reduced regression

Abstract

Diabetes is a chronic disease resulted from lifestyle such as diet. Dietary data is known to have within individual variations. When extracting factors, the large within individual variation can cause problem. We propose to take out the within individual variation in the corvariance (or correlation) matrix before extracting the factors. We will use the 2005-2008 nutritional survey in Taiwan to estimate the within individual variation, then apply this to factor analysis. We will explore the rank reduced regression (RRR) also. Three steps will be taken identifying dietary and lifestyles that associated with high glucose. First, we will use the 1993-1996 Nutrition and Health Survey in Taiwan (NAHSIT 1993-1996) linking the health insurance data and mortality. This will be used to develop the risk prediction models for diabetes. Second, we will use the risk factors, mostly biomarkers, from the first step to find the related dietary and lifestyle factors. In this step, we will use the repeated measurements on intakes to estimate the within individual variation, and use the estimates to adjust the variance-covariance matrix of food intakes. This adjusted matrix will applied to RRR.