Performing Department
STATISTICS
Non Technical Summary
Many statistical procedures rely on an underlying assumption that the data follows a normal distribution -- your standard bell curve type symmetric distribution -- or some other fixed distribution. Having such an assumption gives the statistician a foundation to build off, but it is this foundation -- the assumption of normality or some other distribution -- that is oftentimes flawed. This is where nonparametric statistical methods come into play. As an example, consider the highly engaged topic of crop insurance programs. Under the Clinton administration, the Agricultural Risk Protection Act of 2000 was passed by the legislation increasing the federal budget for the crop insurance programs to $16.1 billion dollars (Public Law 106--224, 2000) over a five year window. Estimating actuarially fair insurance premiums is naturally of great importance both to the providers and users of the insurance programs, however, the insurance premiums the farmers pay are based on
limited historical data and can vary widely based on the model used in the analysis. The limited data size suggests a parametric approach cannot be justified, but a classical nonparametric approach may tend to be too conservative yielding a high premium for the farmers. However, through the use of a special class of nonparametric statistical models, more accurate estimates of crop yield densities can be derived capturing the desired properties of the classical nonparametric estimate yet with a less conservative premium for the farmer. Furthermore, these techniques can be adapted in very general statistical framework. This work will remove the often flawed assumption that the data come from a particular distribution, thus allowing the data to speak for itself. Another limitation frequently encountered is the lack of statistical tools used analyze complex and dynamic data structures. This is particularly so in genetics where genetic data is increasingly available with more and more
complex data structures. In order to analyze such data structures, new statistical techniques must be developed. By advancing the research of such complex data structures, new genetic discoveries can be made ultimately leading to a more productive and resilient plant. Professors George Casella and Rongling Wu in the Department of Statistics here at UF made significant progress in genetic data analysis by incorporating time course data, such as a plant's height over time, into the original statistical genetic framework. Since its inception, this modern technique, termed funtional mapping, has been invoked in a wide array of genetic data structures. However, there is still much more progress to be made, for example, by integrating these techniques into crop modeling. The ultimate value from producing superior crops will be the hopeful elimination of more than 800 million malnourished people throughout the world.
Animal Health Component
(N/A)
Research Effort Categories
Basic
100%
Applied
(N/A)
Developmental
(N/A)
Goals / Objectives
Under this research project, research advances will be made in nonparametric methods, semiparametric methods (which combine together the good qualities of nonparametric and parametric methods), and the functional mapping framework. Listed below are clearly stated goals and objectives of this project: 1) Through the use of a special class of statistical models, more accurate estimates of crop yield densities will be derived capturing the desired properties of the classical nonparametric estimate yet with a less conservative premium for the farmer. 2) Through the integration of crop modeling and functional mapping, new models will be extended to identify pleiotropic QTLs that trigger effects on both vegetative growth and reproductive behavior. 3) Functional mapping will be used to identify the genetic architecture of development-dependent and environmentally sensitive phenotypic traits.
Project Methods
The following steps will be taken to achieve the goals and objectives of this proposal: 1) Nonparametric methods will be further developed and used in applications such as crop yield density estimation. The methods will also be validated with simulation studies. 2) Collaborative research will be conducted with IFAS researchers to integrate functional mapping approach into crop models. The proposed statistical models and algorithms will be employed to analyze the phenotypic and marker data to map the quantitative trait loci underlying reaction norms to predict ecophysiological performances across developmental and environmental contexts. 3) A user-friendly computer implementation of the derived models will be coded and made freely available for all researchers to use.