Research On Robust Tests In Genes Association Studies

With the advances in modern biological technology,there are more than 1000 single nu-cleotide polymorphisms(SNPs)were identified with the association of complex diseases.Firstly,this dissertation reviews most of statistical methods in genome wide association studies from a SNP to multiple SNPs,and introduces some preface knowledge in fields about statistic genetics,probability and mathematical statistic,computational statistic.Secondly,this dissertation proposes 3 robust tests for 3 specific gene association problems,derives the theoretical properties about the 3 robust tests,simulates extensively and applies the 3 robust tests to analyze the real genotype data.The arrangements about the 3 robust tests are as follows:In chapter 3,this dissertation proposes the maximin efficiency robust test(MERT)for multiple nuisance parameters based on theories about the maximin efficiency robust test for only one nuisance parameter and investigates some theoretical properties about this robust test.this dissertation investigates the power of the MERT in a specified scenario intuitively further more and also proposes a meaningful example from statistical genetic field to which the MERT for multiple nuisance parameters can be well applied.Extensive simulation studies are conducted to testify the robustness of the MERT for multiple nuisance parameters and display the good performances of the MERT for multiple nuisance parameters based on the example.In chapter 4,this dissertation proposes a new global test by the use of co-dominant codes for all markers and principal components regression(PCR)theory.The new global test is built on an empirical Bayes type of score statistic for testing marginal associations with each single SNP.The new global test gains power by effectively using linkage disequilibrium among testing SNPs.The new global test reduces to PCR when the genotype for each SNP is coded as the number of minor alleles.This connection lends insight into the power of the new global test relative to PCR and some other popular multi-SNP test.When the real disease causal SNP adopts additive code,the new global test drops some powers with Hardy-Weinberg Equilibrium in the control population."We propose a robust test by taking the minimum p-value of the new global test and PCR based on genotype data adopting additive codes.Through extensive simulation studies and real data analysis about the association between pancreatic cancer and genes,this dissertation shows that the proposed robust test can gain desirable power and can often identify association signals that may be missed by existing tests.Thirdly,multiple outcomes are often collected simultaneously in biomedical fields in or-der to identify whether a continuous response and an ordinal response are associated with a covariate simultaneously.In chapter 5,this dissertation proposes a joint statistical model by the use of a latent variable underlying the ordinal response.Asymptotic results are obtained and a joint test is proposed for testing whether a continuous response or an ordinal response is associated with a covariate simultaneously.Extensive simulations and real data analysis results indicate that the proposed method has more efficient performances than the combined p-values method.Lastly,this dissertation summarizes the works about innovations,shortcomings and merits.This dissertation also points out some future works in the last chapter.