Regresyon modellerinde uç değer incelemede dağılıma bağlı olmayan bir yöntem

---

Thesis Type: Postgraduate

Institution Of The Thesis: Gazi University, Fen Bilimleri Enstitüsü, İSTATİSTİK ANABİLİM DALI, Turkey

Approval Date: 1996

Thesis Language: Turkish

Student: MEHMET YILDIZ

Supervisor: HAMZA GAMGAM

Open Archive Collection: AVESIS Open Access Collection

Abstract:

In the introduction section of this study, several study related with influential value and outliers have been given, and for them the test statistics have been defined. In the first chapter, trimming of outliers and the method of making the estimations depend on the samples which will be selected from the trimmed Bootstrap Population have been explained. Selecting small Bootstrap sample size, mild trimming of the Bootstrap Population and severe trimming of the Bootstrap Population methods can be used to prevent including the outliers in the sample. In this study, trimming of the Bootstrap Population has been interested and also test statistics related with trimming methods have been defined. This method depends on Bootstrap distribution of the estimated residuals. Also, Bootstrap estimations and measure of contamination have been explained. In the second chapter, ah application and a simulation study related with trimming of the Bootstrap Population have been performed. In the application, for a=0.15 trims from Bootstrap Population have been made and trimmed Bootstrap Population has been obtained. In the last chapter, conclusions obtained from this application have been given. Science Code : 406.02.01 Key Words : Outlier, Influential value, Predicted residuals, Bootstrap Page Number : 67 Adviser : Doç. Dr. Hamza GAMGAM n