Automatic arima modelling by the cartesian search algorithm
Authors:
- 1 September 1991
- journal article
- research article
- Published by Wiley in Journal of Forecasting
- Vol. 10 (5) , 465-476
Abstract
In the present study we report on the development and test results of a Cartesian ARIMA Search Algorithm, designed for automatic generation of univariate models for time series data within specified parameter intervals of the identification and estimation stages. Model retention is determined within a preselected set of statistics. By interpreting these statistics as dimensions of the constructed criterion space, we obtain a subset of non‐dominated models according to the rule of maximum dispersion over the efficient set. The CARIMA algorithm allows free specification of the number of criteria used in the runs. The algorithm was tested with both simulated and real economic data. The results based on simulated data indicate that the precision of the CARIMA algorithm is lower for seasonal models and higher for non‐seasonal ones, thus suggesting an inverse relationship between algorithm performance and model complexity.Keywords
This publication has 7 references indexed in Scilit:
- AUTOBOX-PLUS: Version 2.0.Journal of the Royal Statistical Society Series C: Applied Statistics, 1989
- The M-competition with a fully automatic box–jenkins procedureJournal of Forecasting, 1984
- On a measure of lack of fit in time series modelsBiometrika, 1978
- Effective treatment of the interpolation factor in Marquardt's nonlinear least-squares fit algorithmThe Computer Journal, 1975
- An Improved Marquardt Procedure for Nonlinear RegressionsTechnometrics, 1971
- An Analysis of TransformationsJournal of the Royal Statistical Society Series B: Statistical Methodology, 1964
- An Algorithm for Least-Squares Estimation of Nonlinear ParametersJournal of the Society for Industrial and Applied Mathematics, 1963