HUI LIN

We consider the problem of estimating the j oint distribution of two correlated random variables where one is observed with error. An example in nutrition is estimation of the joint distribution of usual energy intake and usual micronutrient intake. While precise biomarkers for energy consumption are available, there are no reliable biomarkers of consumption for nutrients including vitamins and minerals (vitamin K is an exception). Yet, nutritionists are interested in estimating the… Show more

We consider the problem of estimating the j oint distribution of two correlated random variables where one is observed with error. An example in nutrition is estimation of the joint distribution of usual energy intake and usual micronutrient intake. While precise biomarkers for energy consumption are available, there are no reliable biomarkers of consumption for nutrients including vitamins and minerals (vitamin K is an exception). Yet, nutritionists are interested in estimating the distribution of usual intake of micronutrients per unit of caloric intake. This is denoted the nutrient density of the diet and involves estimation of the distribution of the ratio of two non-normal random variables, one of which is observed with measurement error. We develop an approach that combines a deconvolution kernel method (DKM) and the method of copulas to estimate the joint distribution of two non-normal variables where one is contaminated. DKM is first used to adjust the univariate measurement error. A Gaussian copula is then used to model the correlation structure between the two variables after error adjustment. We carried out a small simulation study to investigate whether the two-step method we propose is promising. At least in the context of our simulation, we found that the approach produces good results when the correlation between the two random variables is reasonably high. Our findings are tentative, however, and more research is needed before we can recommend the methodology for use broader. Show less

It was originally motivated by the common practice of pooling diagnostic tests. Matched observations with dichotomous responses commonly occur in medical and epidemiological studies. Although standard approaches exist for one-to-one paired binary data analyses, not much work has been produced for the case where we have one-to-two or one-to-N matched binary data. The existing Miettinen's test assumes that the multiple observations from the same matched set are mutually independent. In this… Show more

It was originally motivated by the common practice of pooling diagnostic tests. Matched observations with dichotomous responses commonly occur in medical and epidemiological studies. Although standard approaches exist for one-to-one paired binary data analyses, not much work has been produced for the case where we have one-to-two or one-to-N matched binary data. The existing Miettinen's test assumes that the multiple observations from the same matched set are mutually independent. In this paper, we propose exact and asymptotic tests for one-to-two matched binary data. Our method is in markedly different from previously proposed methods in that we do not rely on the mutual independence assumption. The emphasis on dependence among observations from the same matched set is natural and appealing, in both human health and in veterinary medicine studies. The method we propose can be applied to many kinds of diagnostic studies that have a one-to-two matched data structure. Our methods can also be generalized to the one-to-N matched case in a straightforward manner. Show less

I was bio-statistician in PADRAP. It is an epidemiologically-based initiative to help producers and veterinarians manage disease risks faced by North American swine industry. It offers a set of risk assessment questionnaires, databases and reports for measuring and benchmarking disease risks. The aim is to identify highly predictive risk factors for clinical outcomes of swine disease based on the survey data.

Undergraduate Creative Research Project Sponsored by Beijing City

Statisitcal methods: Time series analysis including ARCH auto regression, cyclical analysis and spectral analysis

Undergraduate Scientific Research Project Sponsored by Beijing Normal University