Correlation Coefficient Estimation between Incomplete Nutrient Intake Measurements

10.29973/JCSA.200609.0003

Shen-Ming Lee；Shyh-Tyan Ou

Food frequency questionnaire ； Four day food record ； Estimation equation method ； Simple and multiple imputation methods ； Bootstrap sampling

中國統計學報

44卷3期（2006 / 09 / 01）

252 - 270

英文

Nutritional study is important in the epidemiologic research. The food frequency questionnaire, denoted as FFQ, is one of the common instruments for the collection of nutritional data, and is often applied to evaluate the long-term dietary intake. An important issue on the application of FFQ is the validity of FFQ, which can be verified with the degree of the correlation coefficient between the amount of dietary intake measured with FFQ and the amount of long-term dietary intake. Due to the concerns on time-consuming and cost-effective, the two-stage stratified design is often considered for the data collection. The first stage of data collection is to measure initial daily average diet with FFQ from all enrolled subjects. The second stage of data collection, called the calibration sample, is to sample a part of the cohort study. The estimation of the correlation coefficient between FFQ and the long-term dietary intake may be biased if the calibration sample is not a random sample of the study cohort. Comparing to the sampling schemes proposed by Wang, et al. (1999) and Wang (2000), a new sampling scheme is proposed. Meanwhile, three estimation methods proposed for the parameter estimation include the estimation equation method, the simple imputation method and the multiple imputation method. The performances of proposed estimation methods will be investigated in the simulation study. Bootstrap sampling is mainly applied to the variance estimation for estimators of parameters. It is expected that three proposed methods can provide the stable performance on the estimation of parameters in the simulation study.

1. Carroll, R. J.,Freedman, L.,Pee, D.(1997).Design aspects of calibration studies in nutrition, with analysis of missing data in linear measurement error models.Biometrics,53,1440-1457.
2. Freedman, L. S.,Carroll, R. J.,Wax, Y.(1991).Estimating the relationship between dietary intake obtained from a food frequency questionnaire and true average intake.American Journal of Epidemiology,134,310-320.
3. Landin, R.,Freedman, L. S.,Carroll, R. J.(1995).Adjusting for time trends when estimating the relationship between intakes obtained from a food frequency questionnaire and true average intake.Biometrics,51,169-181.
4. Liu, K,Stamler, J,Dyer, A.,McKeever, J.,McKeever, P.(1978).Statistical methods to assess and minimize the role of intra-individual variability in obscuring the relationship between dietary lipids and serum cholesterol.Journal of Chronic Disease,31,399-418.
5. Rubin, D. B.(1987).Multiple Imputation for Nonresponse in Surveys.New York:John Wiley & Sons.
6. Wang, C. Y.(2000).Weighted normality-base estimator in correcting correlation coefficient estimation between incomplete nutrient measurements.Biometrics,56,106-112.
7. Wang, C. Y.,Anderson, G. L.,Prentice, R. L.(1999).Estimation of the correlation between Nutrient intake measures under restricted sampling.Biometrics,55,711-717.