The overall goal of this application is to undertake a detailed statistical examination of caries data obtained from the Iowa Fluoride Study. In this process, novel statistical models will be developed for univariate and bivariate count data both longitudinally, as well as, cross-sectionally and marginally, both at person level and tooth level. In the United States, dental caries this is a major chronic childhood disease. Nevertheless, our understanding of various risk factors is limited. The Iowa Fluoride Study is an ongoing study on a cohort of Iowa children that began in 1991, led by Dr. Steven Levy who is a co-I on this application. It is anticipated that through innovative and efficient statistical modeling, it will e possible to mine the resulting data more fully and discover novel relationships between caries incidences/severity and various potential risk factors. We also plan to reinvestigate the issue of optimal fluoride use through joint modeling of bivariate count data of caries and fluorosis. Prior analysis using univariate approaches indicated that a recommendation of optimal dosage is problematic since the two marginal intervals for fluoride use towards caries and fluorosis prevention did not overlap on the face of subject to subject variability. Thus, the following three broad interconnected aims will be undertaken. We will develop count data models to study the dynamical changes of the various (risk) factors on caries incidence and severity including efficient computational methods for parameter estimation (Aim 1). We will develop various marginal models to understand the relationship between caries incidences and severity with various factors that are applicable across the population and evaluation periods (Aim 2). We will develop bivariate count data regression models to study the joint relationships of caries and fluorosis at comparable evaluations (Aim 3). We will compare our results to those obtained from existing approaches. Statistical software (R packages/codes) for the novel data analysis methods will be freely distributed through the Comprehensive R Archive Network.
The proposed research will lead to a detailed statistical examination of caries data obtained from the Iowa Fluoride Study. It will lead to novel findings regarding the role of potential risk factor for caries. In particular, it will construct a predictie model for the bivariate relationships of development of caries and fluorosis that might lead to new recommendation for the optimal childhood fluoride intake. This research will not only be useful in greater understanding of dental caries but also will it add valuable contribution to biostatistics methodological research of statistical (regression) models for count data.
|Dutta, Sandipan; Datta, Somnath (2016) A rank-sum test for clustered data when the number of subjects in a group within a cluster is informative. Biometrics 72:432-40|
|Bible, Joe; Beck, James D; Datta, Somnath (2016) Cluster adjusted regression for displaced subject data (CARDS): Marginal inference under potentially informative temporal cluster size profiles. Biometrics 72:441-51|
|Choo-Wosoba, Hyoyoung; Levy, Steven M; Datta, Somnath (2016) Marginal regression models for clustered count data based on zero-inflated Conway-Maxwell-Poisson distribution with applications. Biometrics 72:606-18|
|Kong, Maiying; Xu, Sheng; Levy, Steven M et al. (2015) GEE type inference for clustered zero-inflated negative binomial regression with application to dental caries. Comput Stat Data Anal 85:54-66|
|Nevalainen, Jaakko; Datta, Somnath; Oja, Hannu (2014) Inference on the marginal distribution of clustered data with informative cluster size. Stat Pap (Berl) 55:71-92|
|Datta, Somnath; Beck, James D (2014) Robust estimation of marginal regression parameters in clustered data. Stat Modelling 14:489-501|
|Datta, Somnath; Nevalainen, Jaakko; Oja, Hannu (2012) A General Class of Signed Rank Tests for Clustered Data when the Cluster Size is Potentially Informative. J Nonparametr Stat 24:797-808|