Identification of multiple influential observations in Weibull regression

Abdullah, Siti Nabilah Syuhada (2015) Identification of multiple influential observations in Weibull regression. Masters thesis, Universiti Tun Hussein Onn Malaysia.

[img]
Preview
PDF
2580Kb

Abstract

This study focuses on using the Cook's Distance and the Generalised DFFITS in identifling multiple influential observations on a linear regression set up. Non linear distributions are commonly used to model the survival analysis. One such distribution is the Weibull distribution which is used in this study. With the lifetime generated, the performance of Cook's Distance and the Generalised DFFITS in multiple influential observations under percentage of contaminate of 15% and 10%; and various sample sizes of n = 40,50 and 100 are measured. The objectives of the study are to identify multiple influential observations in Weibull Regression and to compare diagnostic method of Cook-type measure and Generalised DFFITS. A simulation study was conducted and the Log-linear form of Weibull distribution was used to generate the lifetime data. The method was also applied to a real data set, namely the patients diagnosed and treated for the human immunodeficiency virus (HIV) between January 1989 until November 1995. On the whole, it is concluded that the Generalised DFFITS diagnostic is the best method of detecting multiple influential observation in a Weibull Regression Model.

Item Type:Thesis (Masters)
Subjects:T Technology > TS Manufactures > TS155-194 Production management. Operations management
ID Code:7503
Deposited By:Normajihan Abd. Rahman
Deposited On:31 Jan 2016 14:51
Last Modified:31 Jan 2016 14:51

Repository Staff Only: item control page