Métodos de estadística robusta
Loading...
Identifiers
Publication date
Authors
Advisors
Tutors
Editors
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
El objetivo de este trabajo es presentar alternativas a los métodos estadísticos clásicos, tanto en el ámbito de la estimación de parámetros como en el de la regresión lineal. Para ello, primero abordaremos el concepto de dato atípico y su detección; así como la introducción al modelo contaminado. Este modelo permitirá un mejor entendimiento del impacto de los datos atípicos en la estimación de parámetros. Además, se definirán diversas medidas de robustez, que nos permitirán comparar los estimadores. En el segundo capítulo, se presentará la estimación robusta de los parámetros unidimensionales de localización y escala mediante el uso de M-estimadores. Se mencionarán sus propiedades, los ejemplos más utilizados y se describirán varios algoritmos para su cálculo. Por último, se introducirán alternativas más robustas al método de regresión de mínimos cuadrados y se aplicarán a ejemplos prácticos para observar su comportamiento. Así, comprenderemos las limitaciones que presenta el método de mínimos cuadrados, ampliamente utilizado, cuando no se cumplen las hipótesis que asume el modelo y las ventajas que presentan las alternativas más robustas.
This project aims to present alternatives to classic statistical methods in the fields of parameter estimation and linear regression. To do so, we will first address the concept of atypical data and its detection, as well as the introduction to the contaminated model. This model will allow for a better understanding of the impact of atypical data in the estimation of parameters. Furthermore, we will define diverse measures of robustness that will allow us to compare the estimators. In the second chapter, we will present the robust estimation of one-dimensional parameters of localization and scale through the use of M-estimators. Their properties and most commonly used examples will be discussed and various algorithms will also be described for their computation. Finally, we will introduce more robust alternatives to the method of least squares regression and we will apply them to practical examples to observe their behavior. Thus, we will understand the limitations of the widely used least squares method when the assumed model hypothesis fail, as well as the advantages of more robust alternatives.
This project aims to present alternatives to classic statistical methods in the fields of parameter estimation and linear regression. To do so, we will first address the concept of atypical data and its detection, as well as the introduction to the contaminated model. This model will allow for a better understanding of the impact of atypical data in the estimation of parameters. Furthermore, we will define diverse measures of robustness that will allow us to compare the estimators. In the second chapter, we will present the robust estimation of one-dimensional parameters of localization and scale through the use of M-estimators. Their properties and most commonly used examples will be discussed and various algorithms will also be described for their computation. Finally, we will introduce more robust alternatives to the method of least squares regression and we will apply them to practical examples to observe their behavior. Thus, we will understand the limitations of the widely used least squares method when the assumed model hypothesis fail, as well as the advantages of more robust alternatives.
Description
61 páxinas
Keywords
Bibliographic citation
Relation
Has part
Has version
Is based on
Is part of
Is referenced by
Is version of
Requires
Sponsors
Rights
Attribution-NonCommercial-ShareAlike 4.0 International







