Variable Selection Of Linear Model With Genotype

The relationship between single genotype and disease has gradually matured in various aspects.However,the traditional study of single genotype and disease neglects the asso-ciation analysis between genes.Because of the particularity of genes,the large number of them and the large amount of information they carry7,it is particularly important to screen the information effectively by introducing the linear model of genotype.In this paper,geno-types AA,Aa and AA are introduced.There are three different genetic models according to different genetic patterns of diseases:invisible model,additive model and dominant model.Three genetic models can be transformed into linear models by transformation.Variable selection is carried out by lasso penalty method.The specific realization of variable selection method is given by random simulation of generated statistical data.The simulation results show that lasso penalty method is feasible in the case of p<n penalty.Because the linear model of introducing genotype has some particularities compared with the ordinary linear model,there are some restrictions on the lasso estimation,which may not be the most ideal variable selection method in the linear model of introducing genotype.This paper uses the naive elastic net method to select the variables of the model,which can be explained by the article.In order to know that lasso-type has defects in the number of variables and group-ing effect,while naive elastic net method has better properties in these aspects under the same computational complexity as lasso-type method,and gives detailed explanation.