In nutritional survey, large within individual variation is common in 24-hr recall data. Therefore, the total variance in nutrition data is large. That complicates the search for dietary pattern of disease. Efforts have been made in identifying dietary patterns for colon cancer. However, the method did not incorporate the large within individual variations. We propose to adjust the large variance by estimating the ratio of within to among individual variation, and to apply the ratio to the variance-covariance matrix for factor analysis. We apply the method to the Nutrition and Health Survey in Taiwan (NAHSIT: 1993-1996) to identify the dietary patterns of hyperurincemia.
Date:
2001-08-08
Relation:
The Joint Statistical Meeting, American Statistical Association. 2001 Aug 8.
