Özyinelemesiz (Non-Recursive) Modellerin Yapısal Analizi ve Sağlık Alanında Bir Uygulama
Abstract
Bu tez alışması, karşılıklı nedensellik ve geri besleme mekanizmaları içeren sistemlerin analizinde özyinelemesiz (non-recursive) yapısal eşitlik modellerinin sunduğu yöntemsel olanakları inceleyen istatistik odaklı bir araştırma olarak tasarlanmıştır. Geleneksel özyinelemeli (recursive) modeller, değişkenler arasındaki ilişkileri çoğunlukla tek yönlü varsayımlar altında ele almakta; bu durum eşzamanlı ve çift yönlü etkileşimlerin bulunduğu yapılarda önemli sınırlılıklar doğurmaktadır. Bu tezde, söz konusu sınırlılıkları aşmak amacıyla non-recursive yapısal eşitlik modelleme yaklaşımı benimsenmiş ve modelin tanımlanabilirliği, kararlılığı ve uyum özellikleri ayrıntılı biçimde değerlendirilmiştir. Araştırma kapsamında kurulan model, birden fazla gizil değişken arasında eşzamanlı ve karşılıklı ilişkileri içeren bir yapıda kurgulanmış; ölçüm ve yapısal bileşenler birlikte ele alınarak iki aşamalı yapısal eşitlik modelleme yaklaşımı izlenmiştir. Ölçüm modeli doğrulayıcı faktör analizi ile test edilmiş; gizil değişkenlerin gözlenen göstergeler tarafından yeterli düzeyde temsil edildiği gösterilmiştir. Yapısal modelde ise özellikle iki gizil yapı arasındaki çift yönlü ilişki, özyinelemesiz bir çerçevede tanımlanarak geri beslemeli bir sistem kurulmuştur. Non-recursive modellemenin gerektirdiği tanımlanabilirlik koşulları doğrultusunda, modelde gerekli yapısal kısıtlamalar uygulanmış ve serbestlik derecesi pozitif olacak biçimde yapılandırma gerçekleştirilmiştir. Parametre tahminleri maksimum olabilirlik yöntemiyle elde edilmiş; modelin çözüm kararlılığı ve istatistiksel tutarlılığı uyum iyiliği indeksleri ve stabilite ölçütleri üzerinden değerlendirilmiştir. Elde edilen bulgular, kurulan modelin kabul edilebilir uyum düzeylerine sahip olduğunu ve karşılıklı nedensellik içeren ilişkilerin tek bir analitik çerçevede güvenilir biçimde tahmin edilebildiğini göstermektedir. Bu çalışma, sağlık alanından elde edilen ikincil bir veri seti üzerinde yürütülmüş olmakla birlikte, temel katkısını kullanılan veriden ziyade uygulanan istatistiksel modelleme yaklaşımından almaktadır. Özellikle, non-recursive yapısal eşitlik modellerinin sağlık bilimleri bağlamında uygulanabilirliğinin gösterilmesi, model tanımlanabilirliği ve yapısal çözüm sürecine ilişkin metodolojik kararların açık biçimde ortaya konulması, çalışmanın yöntemsel özgünlüğünü oluşturmaktadır. Sonuç olarak bu tez, karşılıklı etkileşim ve geri besleme içeren karmaşık sistemlerin analizinde özyinelemesiz yapısal eşitlik modellerinin klasik yöntemlere kıyasla sağladığı analitik üstünlükleri ortaya koymakta; biyoistatistik alanında bu modellerin uygulanmasına yönelik kapsamlı ve örnekleyici bir metodolojik çerçeve sunmaktadır. Anahtar kelimeler: Doktor Hasta Etkileşimi, Hasta Memnuniyeti, Özyinelemesiz Modeller, Yapısal Eşitlik Modellemesi
This thesis study is designed as a statistical-oriented research that examines the methodological possibilities offered by non-recursive structural equation models in the analysis of systems involving reciprocal causality and feedback mechanisms. Traditional recursive models mostly consider the relationships between variables under one-way assumptions; This situation creates significant limitations in structures with simultaneous and bidirectional interactions. In this thesis, a non-recursive structural equation modeling approach was adopted in order to overcome these limitations and the definability, stability and fit properties of the model were evaluated in detail. The model established within the scope of the research is designed in a structure that includes simultaneous and reciprocal relationships between more than one latent variable; measurement and structural components were considered together and a two-stage structural equation modeling approach was followed. The measurement model was tested with confirmatory factor analysis; It has been shown that latent variables are adequately represented by the observed indicators. In the structural model, the bidirectional relationship between two latent structures is defined in a non-recursive framework and a feedback system is established. In line with the identifiability conditions required by non-recursive modeling, the necessary structural constraints were applied in the model and the configuration was carried out in a way that the degree of freedom was positive. Parameter estimates were obtained by the maximum likelihood method; The solution stability and statistical consistency of the model were evaluated through goodness-of-fit indices and stability criteria. The findings show that the established model has acceptable levels of agreement and that mutually causal relationships can be reliably predicted in a single analytical framework. Although this study was conducted on a secondary dataset obtained from the field of health, its main contribution is derived from the statistical modeling approach applied rather than the data used. In particular, demonstrating the applicability of non-recursive structural equation models in the context of health sciences, model identifiability and clear methodological decisions regarding the structural solution process constitute the methodological originality of the study. As a result, this thesis reveals the analytical advantages of non-recursive structural equation models compared to classical methods in the analysis of complex systems involving mutual interaction and feedback; It provides a comprehensive and exemplary methodological framework for the application of these models in the field of biostatistics. Keywords: Doctor-Patient Interaction, Patient Satisfaction, Non-Recursive Models, Structural Equation Modeling,
This thesis study is designed as a statistical-oriented research that examines the methodological possibilities offered by non-recursive structural equation models in the analysis of systems involving reciprocal causality and feedback mechanisms. Traditional recursive models mostly consider the relationships between variables under one-way assumptions; This situation creates significant limitations in structures with simultaneous and bidirectional interactions. In this thesis, a non-recursive structural equation modeling approach was adopted in order to overcome these limitations and the definability, stability and fit properties of the model were evaluated in detail. The model established within the scope of the research is designed in a structure that includes simultaneous and reciprocal relationships between more than one latent variable; measurement and structural components were considered together and a two-stage structural equation modeling approach was followed. The measurement model was tested with confirmatory factor analysis; It has been shown that latent variables are adequately represented by the observed indicators. In the structural model, the bidirectional relationship between two latent structures is defined in a non-recursive framework and a feedback system is established. In line with the identifiability conditions required by non-recursive modeling, the necessary structural constraints were applied in the model and the configuration was carried out in a way that the degree of freedom was positive. Parameter estimates were obtained by the maximum likelihood method; The solution stability and statistical consistency of the model were evaluated through goodness-of-fit indices and stability criteria. The findings show that the established model has acceptable levels of agreement and that mutually causal relationships can be reliably predicted in a single analytical framework. Although this study was conducted on a secondary dataset obtained from the field of health, its main contribution is derived from the statistical modeling approach applied rather than the data used. In particular, demonstrating the applicability of non-recursive structural equation models in the context of health sciences, model identifiability and clear methodological decisions regarding the structural solution process constitute the methodological originality of the study. As a result, this thesis reveals the analytical advantages of non-recursive structural equation models compared to classical methods in the analysis of complex systems involving mutual interaction and feedback; It provides a comprehensive and exemplary methodological framework for the application of these models in the field of biostatistics. Keywords: Doctor-Patient Interaction, Patient Satisfaction, Non-Recursive Models, Structural Equation Modeling,
Description
Keywords
Biyoistatistik, Biostatistics
Turkish CoHE Thesis Center URL
WoS Q
Scopus Q
Source
Volume
Issue
Start Page
End Page
59

