Variable Selection Of Genetic Model In Longitudinal Data

Over a period of time,observations of repeated measurements of different individuals are called longitudinal data.Longitudinal data has a wide range of applications in medicine and biology.For example,in medicine,the data collected by patients during their follow-up visits is the longitudinal data.At the same time,compared with cross-section data in gene association analysis,longitudinal data is compared with cross-sectional data in gene association analysis.It can study the relationship between the individual's own disease complexity and changes over time,so it can improve the genetic variation of complex diseases.In recent years,in the study of genes and longitudinal data,the main focus has been on the study of the impact of genome association analysis on individual traits in the form of longitudinal data.For example,Wang et al(2012)and Inan G(2017)mentioned the use of longitudinal data models to study gene association analysis.The former uses the longitudinal data model to research the gene expression level of yeast cells in a cycle or several cycles;the latter uses the longitudinal data model to identify transcription factors that regulate the gene expression level of cells during the cycle;however,in the above studies It does not take into account the uncertainty of the genetic model.If the uncertainty of the genetic model is not considered in the modeling,it will have a certain impact on our statistical inference.Therefore,this article mainly discusses the consideration of the uncertain situation of the genetic model at the gene locus in the genetic association analysis of longitudinal data,combined with the idea of punishment,gives the method of variable selection such as lasso and PGEE and finally obtains the parameter estimation.Finally Numerical simulation is used to illustrate the rationality of the variable selection method.