|
Activity Number:
|
310
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Wednesday, August 14, 2002 : 10:30 AM to 12:20 PM
|
|
Sponsor:
|
Business & Economics Statistics Section*
|
| Abstract - #301820 |
|
Title:
|
Flexible Semiparametric Forecasting
|
|
Author(s):
|
Dimitrios Thomakos*+
|
|
Affiliation(s):
|
Florida International University
|
|
Address:
|
DM307B, University Park Campus, Miami, Florida, 33199, USA
|
|
Keywords:
|
Flexible ; Semiparametric ; Forecasting ; Exponential Smoothing ; Functional Coefficients ; Nonlinearity
|
|
Abstract:
|
I propose a method for time series forecasting that combines two existing approaches (exponential smoothing and functional coefficients) into a single model. The model allows for modelling flexibility due to: (a) the limited set of assumptions placed on the, non-parametric, functional form that the model coefficients can take; (b) the semi-parametric method of estimation of that functional form; and (c) the standard methods of estimation of the complete model. The proposed method can be applied into several forecasting settings. It allows for non-linearity of generic, non-parametric form in either the dependent variable or the transfer variable. It allows for adaptive weighting between a linear component and a non-linear one. The weights can be made time-dependent, and therefore provide additional flexibility for transitions between the two components. Estimation can be carried out using an iterative approach that can involve one or more of linear least squares, functional approximations and non-parametric regression. Obtaining one-step ahead forecasts is straightforward. Multi-step ahead and interval forecasts can be computed by an application of the bootstrap.
|
- The address information is for the authors that have a + after their name.
- Authors who are presenting talks have a * after their name.
Back to the full JSM 2002 program |
