RESEARCH IN BIOSTATISTICAL METHODOLOGY AND MATHEMATICAL MODELING

不同的统计方法调整宏量营养素摄入的效果

Different statistical methods that adjust the effect of macronutrient intak for total energy intake are currently being used to analyze epidemiologic studies of diet and disease. This research project is examining the statistical properties of these methods. During the past year this work resulted in 1) a paper published on the ability of currently-used energy adjustment regression methods to disentangle the effects of total energy from its component macronutrient-specific parts, and 2) a paper conditionally accepted for publication that investigates the effect of dietary measurement error on the results of applying energy adjusted methods for a typical case of measurement errors that are heteroscedastic, non-normal, and correlate with nutrient intakes and with each other. Two projects were undertaken to improve standard confidence limits for estimating age-specific and age-adjusted cancer incidence and mortality rates. The first project investigated methods of generalizing standard Poisson variance methods to allow extra-Poisson variation in the calculation of confidence limits. We investigated a nonparametric nearest-neighbor estimate of the variance which does not need to rely on any assumed regression model and found it to give an approximately unbiased estimate of the overdispersion parameter for a wide class of true regression models. Once the overdispersion is nonparametrically estimated, the resulting variance estimate provides a more robust confidence interval estimate of the cancer rates. The second project investigated confidence intervals for age-adjusted rates under the Poisson assumption. We found that the normal approximations as well as a recently proposed approximation do not perform well when the number of counts is small and the adjustment weights vary substantially across the different ages. We propose an approximation which gives exact intervals whenever the age-adjusted rate reduces to a weighted Poisson random variable. For other cases we compare our approximation to other methods by simulations and show that it has better coverage properties. In collaboration with Dr. Chris Gennings of the Medical College of Virginia this project has developed a nonparametric permutation test of estimated distribution functions for comparing two groups of individuals, each having a set of repeated measurements on an ordinal scale. This methodology is broad enough to be applied to right censored and interval censored survival data. For right censored data, this ridit permutation approach leads to a known rank test. New statistical methods for detection and inference of disease clusters are being developed. Previously proposed methods have been tests for overall clustering and do not have the ability to identify the location of clusters. The properties of a spatial scan statistic which takes into account the nonhomogeneous population densities as well as confounding variables have been evaluated. The statistic both tests for the location o clusters and tests for their statistical significance, and thus can be used both to evaluate cluster alarms and for routine surveillance. The spatial scan statistic has also been extended to study space-time clusters in addition to purely spatial ones.