Type
Text
Type
Dissertation
Advisor
wu, song | zhu, wei | yang, jie | DeLorenzo, Christine.
Date
2017-05-01
Keywords
Statistics
Department
Department of Applied Mathematics and Statistics
Language
en_US
Source
This work is sponsored by the Stony Brook University Graduate School in compliance with the requirements for completion of degree.
Identifier
http://hdl.handle.net/11401/77266
Publisher
The Graduate School, Stony Brook University: Stony Brook, NY.
Format
application/pdf
Abstract
Interval and linkage mapping are currently the most popular approaches for QTL mapping method. If phenotypic traits of interest are continuous, they are often assumed to follow a Gaussian mixture model. In this way, standard ML approach and LR test can be used to find the estimates of parameters and the position of a QTL. However, the ML approach cannot be applied appropriately under the case of heterogeneous variances, due to the singularities of the likelihood function. In order to solve the problem of degeneracy, we derived a suitable penalty function to penalize the likelihood function. It can allow heterogeneous variances in the Gaussian mixture model and the test of the presence of single QTL in a genome. Extensive simulation studies have been performed to compare the penalized method with standard ML approach on the power of detecting the existence of QTL. Our results demonstrate that under the scenario of heterogeneous variance, the penalized method outperforms the unpenalized one in power and provide a robust estimation of the model. | 87 pages
Recommended Citation
MENG, ZIQI, "Penalization for Gaussian mixture model and its application" (2017). Stony Brook Theses and Dissertations Collection, 2006-2020 (closed to submissions). 3087.
https://commons.library.stonybrook.edu/stony-brook-theses-and-dissertations-collection/3087