Publication: Genelleştirilmiş Lineer Modellerde Tanısal Grafikler
Loading...
Date
Authors
Journal Title
Journal ISSN
Volume Title
Abstract
Genelleştirilmiş lineer modeller fen ve mühendislik alanlarında sıkça karşılaşılan kesikli veya sürekli verilerin analizine olanak veren modellerden biridir. Lineer modellerde bağımlı değişken ile bağımsız değişkenler arasındaki ilişkinin doğru biçimde belirlenmesinde kullanılan çeşitli istatistiksel ve tanısal grafik yöntemler bulunmaktadır. Bu çalışmada bu yöntemlerin Genelleştirilmiş Lineer Modellerde de uygulanabilirliliği incelenmiştir. Genelleştirilmiş Lineer Modellerde tanısal grafik yöntemlere bağımlı değişken ile bağımsız değişkenleri ilişkilendiren bağıntı fonksiyonunun doğru belirlenip belirlenmediğini özellikle bağıntı fonksiyonunda yer alan bağımsız değişkenlerdeki ilişkinin biçimini ortaya çıkarmak için başvurulduğu gösterilmiştir. Bu amaçla uygulanan bazı tanısal grafik yöntemlerin bağımsız değişkenler arasında mevcut olabilecek çoklu birlikte değişim probleminden dolayı doğru sonuçlar vermeyeceği gösterilmiştir. Bu duruma çoklu bağlantı problemi olmayan ve olan iki sayısal veri türetilerek değinilmiştir. Ayrıca 1965-1998 yılları arasındaki grev sayılarını sendikalaşma oranı, ücret oranı işyeri ve kapanan işyeri sayıları ile analiz ederek en uygun model seçilmeye çalışılmıştır.
Generalized Linear Models provide opportunity to analyse the data obtained from Science and Engineering, which have discrete or continued distribution. In this thesis Generalized Linear Models was studied. The statistical and diagnostic graphical methods were considered to detect if the relationship between dependent variable and independent variables is identified correctly or not. There are several techniques in Lineer Models to detect the right form of the relationship between dependent variable and independent variables. In this study we pointed out that these techniques can be applied in Generalized Linear Models. It was demonstrated that the diagnostics graphs in Generalized Linear Models are applied to reveal the true form of the link function which connects the dependent variable and the independent variables and, especially to identify nonlinearity in covariates used in the link function . It was presented that some of the graphical methods applied for this purpose can be failed when there is multicolinearity problem in covariates. To illustrate this we generated two data sets without multicolinearity problem and with it, respectively. The number of the strike in the years 1965-1998 was tried to explain with the rate of the inclusion of a labour union, the number of labours, the rate of wages and the number of shut-down labours and. It was tried to find a best model for this relationship.
Generalized Linear Models provide opportunity to analyse the data obtained from Science and Engineering, which have discrete or continued distribution. In this thesis Generalized Linear Models was studied. The statistical and diagnostic graphical methods were considered to detect if the relationship between dependent variable and independent variables is identified correctly or not. There are several techniques in Lineer Models to detect the right form of the relationship between dependent variable and independent variables. In this study we pointed out that these techniques can be applied in Generalized Linear Models. It was demonstrated that the diagnostics graphs in Generalized Linear Models are applied to reveal the true form of the link function which connects the dependent variable and the independent variables and, especially to identify nonlinearity in covariates used in the link function . It was presented that some of the graphical methods applied for this purpose can be failed when there is multicolinearity problem in covariates. To illustrate this we generated two data sets without multicolinearity problem and with it, respectively. The number of the strike in the years 1965-1998 was tried to explain with the rate of the inclusion of a labour union, the number of labours, the rate of wages and the number of shut-down labours and. It was tried to find a best model for this relationship.
Description
Tez (yüksek lisans) -- Ondokuz Mayıs Üniversitesi, 2005
Libra Kayıt No: 16933
Libra Kayıt No: 16933
Keywords
Citation
WoS Q
Scopus Q
Source
Volume
Issue
Start Page
End Page
98
