Modelos de regresión aditivos
Loading...
Identifiers
Publication date
Authors
Advisors
Tutors
Editors
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
En muchas situaciones es de interés poder representar la relación de dependencia entre una variable respuesta y una o varias variables explicativas. Con este propósito introducimos los modelos de regresión. En una primera aproximación, lo más intuitivo es plantear un modelo de regresión paramétrico, es decir, que la forma del modelo sea totalmente conocida salvo por un cierto vector de parámetros, como es el caso de los modelos de regresión lineales. Sin embargo, en la práctica estos modelos no siempre ajustan bien la relación entre las variables que queremos representar, por lo que debemos recurrir a otro tipo de relaciones que nos den una mayor flexibilidad. En este contexto proponemos los modelos de regresión aditivos. Los modelos de regresión aditivos son modelos no paramétricos, por lo que son muy flexibles, pero a la vez su formulación nos permite interpretar el efecto que tiene cada una de las variables explicativas sobre la variable respuesta. A lo largo de este manuscrito se presentarán métodos de estimación de los modelos aditivos (usando ideas de mínimos cuadrados) prestando especial interés a la elección de los parámetros de suavizado, como en cualquier modelo no paramétrico. Finalmente, se ilustrará la utilidad de los modelos aditivos empleando una base de datos reales
In many situations it is interesting to be able to represent the dependence relationship between a response variable and one or several explanatory variables. With this purpose we introduce regression models. In a first approximation, the most intuitive approach is to formulate a para- metric regression model, that is, the form of the model would be completely known except for a certain parameter vector, as in the case of linear regression models. However, in practice these models not always adjust well the relationship between the variables we want to represent, so we must turn to other types of relationships which give us more flexibility. In this context we propose additive regression models. Additive regression models are non-parametric models, so they are very flexible, but at the same time its formulation allows us to interpret the effect that each one of the explanatory variables has on the response variable. Throughout this manuscript estimation methods for additive models will be presented (using least squares ideas) paying special interest to the choice of the smoothing parameters, as in any non-parametric model. Finally, the utility of additive models will be ilustrated using a real database.
In many situations it is interesting to be able to represent the dependence relationship between a response variable and one or several explanatory variables. With this purpose we introduce regression models. In a first approximation, the most intuitive approach is to formulate a para- metric regression model, that is, the form of the model would be completely known except for a certain parameter vector, as in the case of linear regression models. However, in practice these models not always adjust well the relationship between the variables we want to represent, so we must turn to other types of relationships which give us more flexibility. In this context we propose additive regression models. Additive regression models are non-parametric models, so they are very flexible, but at the same time its formulation allows us to interpret the effect that each one of the explanatory variables has on the response variable. Throughout this manuscript estimation methods for additive models will be presented (using least squares ideas) paying special interest to the choice of the smoothing parameters, as in any non-parametric model. Finally, the utility of additive models will be ilustrated using a real database.
Description
70 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








