QA547 : On Variable Selection Methods in High Dimensional Regression Models
Thesis > Central Library of Shahrood University > Mathematical Sciences > MSc > 2019
Authors:
Abstarct: Nowadays, due to the development of technology, it has been possible to store and analyze data with a large number of variables: but it should be noted that if the recorded variables are not significant or the number of them is high, classical estimation techniques are ineffective and in addition to identifying and removing redundant variables, non-classical methods for estimating model parameters should be used. In the regression modeling, by using penalized estimators, we are able to select both significant variables and estimate parameters. In this regard, penalized group estimators can be used if the variables are grouped in such a way that all of them have to be eliminated or they should all remain in the model. In this dissertation, we briefly review some of the most common and applicable modern variable selection methods and investigate penalized group estimators via a series of numerical studies, simulations, and analysis of genomic real data and their use in logistic regression models.
Keywords:
#Genomics data #Group LASSO #High dimensional #SCAD #Variable selection. Link
Keeping place: Central Library of Shahrood University
Visitor:
Keeping place: Central Library of Shahrood University
Visitor: