Grant: Discovery Grants Program - Individual

Dynamic models describe the rate of change of a process defined in time, space or some other continuum. They are widely used in medicine, engineering, ecology and a host of other applications. A central but difficult problem is estimation of dynamic models from observations. We propose a novel solution with several unique aspects. Importantly, the computational load is much lower than traditional methods, and the new method can be applied in situations where observations are missing for some components in the dynamic system. This proposal will consider two specific and important application areas as described below. Many HIV dynamic models have been proposed to describe dynamic interactions between uninfected cells, infected cells and the HIV virus. I will focus on dynamic model selection from noisy data. Biologists have known that each gene contains a particular set of instructions that usually code for a particular protein. Some genes interact with each other and compose a gene regulatory network. I will identify dynamic models for gene regulatory networks. I also propose a new method to determine the effects of air pollutants on public health. The statistical methodology currently available may overestimate effects of air pollutants on public health and underestimate their precision. I will develop a general method which addresses these difficulties. Additionally, software for these methods will be made publicly available.