5128

4818

كارشناسي ارشد

آمار﴿ اقتصادي - اجتماعي﴾

اصفهان: دانشگاه صنعتي اصفهان،دانشكده رياضي

1388

[دوازده]،127ص.:مصور،جدول،نمودار

ص.ع.به فارسي و انگليسي

سروش عليمرادي،علي رجالي

30/1/89

عبدالرسول برهاني حقيقي،علي زينل همداني

1396/09/28

كتابنامه

رياضي

ID4818

به فارسي و انگليسي:قابل رويت در نسخه ديجيتالي

Multicollinearity and Logistic Regression Marzieh Shahmandi Honejani m shahmandihonejani@math iut ac ir february 17 2010 Master of Science Thesis in Farsi Department of Mathematical Sciences Isfahan University of Technology Isfahan 84156 83111 IranSupervisor Dr Soroush Alimoradi salimora@cc iut ac irSupervisor Dr Ali Rejali a rejali@cc iut ac ir2000 MSC 62H35 Key words Multiple Logistic Regression Multicollinearity Principal Component Partial Least SquaresRegression Bootstrap Stein and Ridge Estimators AbstractIn this thesis we present multicollinearity in logistic regression based on an article by Aguil era et al 2006 Multiple logistic regression model is used to predict a binary variable interms of a set of explanatory ones This regression with correlated explanatory variables is relevant to a broad range of prob lems in the physical chemical and enginearing sciences In multiple logistic regression ifthe explanatory variables are dependent it seems that no one variable is important whenall the other are in the model then we have unstable model and the estimated parametersare inaccurate So the interpretation of the relationships between the response and eachexplanatory variable in terms of odds ratios may have some errors There are some methodsto overcome this problem One of these methods using a class of principal component estimators for logistic regression de ned by a scaling parameter By this additional parameter we introduce a class of stan dardized explanatory variables The other one is generalized partial least squares method for logistic regression this methodde nes incorrelated variables P LS components given by linear spans of original explanatoryvariables and uses them as new explanatory variables of the regression model Another method for decreasing e ects of multicollinearity in logistic regression is Ridge andStein estimation methods In this thesis we are going to introduce and apply these methods to di erent data sets In chapter one we have a history and introduction to the subject Chapter two deals withintroducing the necessary de nitions and concepts Then in Chapter three we present thede nition of partial least squares regression and partial least squares generalized linear regres sion and explains the detail of these methods In chapter four we study a class of principalcomponent Ridge and Stein estimators in the logistic regression and in chapter ve we con tinue with an applications of the methods In chapter six the results will be explained

سروش عليمرادي،علي رجالي

عبدالرسول برهاني حقيقي،علي زينل همداني