| |
 |
|
| |
|
|
Modelos no paramétricos en la determinación del spread en un mercado primario de renta fija
María Bonilla Musoles, Leandro García Menéndez, MªLuisa Martí Selva and Rosa Puertas Medina |
|
The recent proposal of various non-parametric models to solve prediction problems in banking
has provided powerful estimators. This paper uses two non parametric models, Classification and
Regression Trees (CART) and Multivariate Adaptative Regression Splines (MARS), in the eurobond
market to measure the accuracy of both these methods in determining issue price in the primary market.
Also the research included a comparative analysis with a parametric model, lineal regression estimated
by Ordinary Least Squares (OLS). Furthermore, the learning-generalization dilemma is tackled
by cross-validation in order to establish the most efficient algorithm structure for the solution of our
problem. The study concludes that CART algorithm is best at banking and state sector spread prediction,
while MARS algorithm and OLS estimates are more accurate at banking issues.
|
|
| Copyright
1998 - 2012 - Asociación Española de Finanzas (AEFIN).
Todos los derechos reservados
e-mail: aefin@aefin.es
y secretaria@aefin.es
c/ Rosario Pino, 8, 10º-A
28020 Madrid
Tlf. y fax:
+34 91 3691483 |
|
