The ongoing growth of scientific production across all fields of knowledge poses a fundamental challenge: how can we synthesize the results of various published studies to draw more accurate conclusions? Meta-analysis addresses the need to combine the results of multiple studies, but the heterogeneity often present between them complicates inference. This work focuses on the meta-analysis of odds ratios. The objective is to examine different statistical approaches to account for between-study heterogeneity, comparing classical fixed-effect and random-effects models with more recent alternatives such as beta-binomial models. Particular attention is paid to how the latter model the presence of overdispersion, introducing different methods for estimating the intra-class correlation parameter. Finally, through a simulation study, the performance of the different approaches is compared to provide a comparative overview useful for selecting the most appropriate model.
La continua crescita della produzione scientifica in ogni ramo dello scibile pone una sfida fondamentale: come sintetizzare i risultati di diversi studi pubblicati per trarre conclusioni più accurate? La meta-analisi risponde alla necessità di combinare i risultati di più studi, ma l’eterogeneità, spesso presente tra questi, complica l’inferenza. Questo lavoro si concentra sulla meta-analisi per rapporti di quote. L’obiettivo è esaminare diversi approcci statistici per tener conto dell’eterogeneità tra gli studi, confrontando i modelli classici ad effetti fissi e casuali con alternative più recenti come i modelli beta-binomiali. Viene posta particolare attenzione a come questi ultimi modellano la presenza di sovradispersione, introducendo diversi metodi per la stima del parametro di correlazione intra-classe. Infine, tramite uno studio di simulazione, si confronta l’efficacia dei diversi approcci per fornire una panoramica comparativa utile alla scelta del modello più adeguato.
Modelli statistici per la meta-analisi di rapporti di quote
DE PIETRO, LEONARDO
2024/2025
Abstract
The ongoing growth of scientific production across all fields of knowledge poses a fundamental challenge: how can we synthesize the results of various published studies to draw more accurate conclusions? Meta-analysis addresses the need to combine the results of multiple studies, but the heterogeneity often present between them complicates inference. This work focuses on the meta-analysis of odds ratios. The objective is to examine different statistical approaches to account for between-study heterogeneity, comparing classical fixed-effect and random-effects models with more recent alternatives such as beta-binomial models. Particular attention is paid to how the latter model the presence of overdispersion, introducing different methods for estimating the intra-class correlation parameter. Finally, through a simulation study, the performance of the different approaches is compared to provide a comparative overview useful for selecting the most appropriate model.| File | Dimensione | Formato | |
|---|---|---|---|
|
De_Pietro_Leonardo.pdf
accesso aperto
Dimensione
10.47 MB
Formato
Adobe PDF
|
10.47 MB | Adobe PDF | Visualizza/Apri |
The text of this website © Università degli studi di Padova. Full Text are published under a non-exclusive license. Metadata are under a CC0 License
https://hdl.handle.net/20.500.12608/88519