Yarı Parametrik Regresyon Modelinde Etkili Gözlem Analizi

Abstract

The aim of thesis is to develope diagnostics, used to detect influential observations in linear regression model, for semiparametric regression model and examine whether the developed diagnostics are successful or not in determining influential observations using a real data set and simulation. In this study, methods used for estimation of parameters in semiparametric regression are introduced, then, diagnostics such as Cook’s Distance, Hadi’s diagnostic, Pena’s diagnostic and COVRATIO which are frequently used in recent years are examined and Hadi’s diagnostic, Pena’s diagnostic and COVRATIO are developed for semiparametric regression model. It is investigated whether developed diagnostics detect or not the influential observations by using real data and artificial data. It is examined and compared the success of developed diagnostics in determining influential observations in different sample sizes via simulation study.