Linear regression: an alternative to logistic regression through the non-parametric regression
For applying the logistic regression, and any other type of parametric regression method, it is necessary to know the model that we want to adjust. In the case of a logistic regression, with only one independent variable, the use of a non- parametric regression is useful in order to obtain evidence...
Saved in:
Published in | Investigación operacional Vol. 38; no. 3; p. 247 |
---|---|
Main Authors | , , , |
Format | Journal Article |
Language | English |
Published |
Editorial Universitaria de la Republica de Cuba
01.09.2017
|
Subjects | |
Online Access | Get full text |
Cover
Loading…
Summary: | For applying the logistic regression, and any other type of parametric regression method, it is necessary to know the model that we want to adjust. In the case of a logistic regression, with only one independent variable, the use of a non- parametric regression is useful in order to obtain evidence of the possible relation between the success probability (n) and the independent variable (x). With this information is possible to determine the model to be adjusted. In this work, it is proposed to use a linear regression model, where the dependent variable is the probability (n), and whose values are obtained from the estimated probabilities that resulting in the application of non-parametric regression. This proposal would avoid using logistic regression, whenever it is necessary to apply a non-parametric regression, for obtaining information of the type of model to be considered. KEYWORDS: Linear regression, logistic regression, non-parametric regression. MSC: 62G08; 62J12 Para la aplicacion de la regresion logistica y cualquier otro tipo de regresion parametrica, es necesario conocer el modelo que se desea ajustar. En el caso de una regresion logistica, con una sola variable independiente, el uso de una regresion no parametrica es de utilidad para obtener evidencia de la posible relacion entre la probabilidad de exito (n) y la variable independiente (x). Con esta informacion es posible determinar el modelo que debe ajustarse. En este trabajo se propone usar un modelo de regresion lineal, donde la variable dependiente es la probabilidad (X), cuyos valores son obtenidos de las probabilidades estimadas que resultan de la aplicacion de la regresion no parametrica. Esta propuesta evitaria usar la regresion logistica, siempre que sea necesario aplicar una regresion no parametrica para obtener informacion del tipo de modelo a considerar. |
---|---|
ISSN: | 0257-4306 |