Extension and implementation of bootstrap-based approaches for inference on the total deviation index
The total deviation index (TDI) is an unscaled statistical measure used to evaluate the deviation between paired quantitative measurements when assessing agreement between different raters. It describes a boundary such that a large specified proportion of the differences in paired-measurements are w...
| Autor: | |
|---|---|
| Tipo de recurso: | tesis de maestría |
| Fecha de publicación: | 2025 |
| País: | España |
| Institución: | Universitat Politècnica de Catalunya (UPC) |
| Repositorio: | UPCommons. Portal del coneixement obert de la UPC |
| Idioma: | inglés |
| OAI Identifier: | oai:upcommons.upc.edu:2117/441458 |
| Acceso en línea: | https://hdl.handle.net/2117/441458 |
| Access Level: | acceso embargado |
| Palabra clave: | Mathematical statistics R (Computer program language) agreement bootstrap simulation study total deviation index Estadística matemàtica R (Llenguatge de programació) Classificació AMS::62 Statistics::62P Applications Classificació AMS::62 Statistics::62G Nonparametric inference Àrees temàtiques de la UPC::Matemàtiques i estadística |
| Sumario: | The total deviation index (TDI) is an unscaled statistical measure used to evaluate the deviation between paired quantitative measurements when assessing agreement between different raters. It describes a boundary such that a large specified proportion of the differences in paired-measurements are within the boundary. The inference of the TDI involves the estimation of a 100(1-α)% upper bound, where α is the significance level. Already existing methodologies for TDI estimation and inference include both parametric (Choudhary, Escaramís et al.) and non-parametric approaches (Choudhary, Perez-Jaume and Carrasco). In this work, the goals are to propose and evaluate new bootstrap-based approaches for inference on the TDI in the method of Perez-Jaume and Carrasco and to implement the existing and new methodologies in an R package. We apply all methodologies to two real-world datasets with different distributions and we also conduct a simulation study to assess their performance. We conclude that under normally-distributed data the parametric methods perform better but under skewed data agreement should be assessed via the non-parametric methods. |
|---|