-
Improved spread-location visualization
Authors:
A. Ian McLeod
Abstract:
The spread-location plot has often been used as a diagnostic plot suitable for many types of fitted statistical models. The spread-location plot which plots the absolute residual or square-root absolute residual versus fitted value along with a robust loess smooth is a useful replacement for the customary practice of plotting residuals versus fitted values. In this note, we show that neither absol…
▽ More
The spread-location plot has often been used as a diagnostic plot suitable for many types of fitted statistical models. The spread-location plot which plots the absolute residual or square-root absolute residual versus fitted value along with a robust loess smooth is a useful replacement for the customary practice of plotting residuals versus fitted values. In this note, we show that neither absolute residual or square-root absolute residual is always appropriate for error distributions likely to be encountered in actual applications. Hence we recommend a multipanel display showing a suitable transformation of the absolute residual versus fitted value along with a boxplot to judge the symmetry achieved by the transformation. We conclude with an illustrative example.
△ Less
Submitted 4 November, 2016;
originally announced November 2016.
-
Parsimony, model adequacy and periodic correlation in forecasting time series
Authors:
A. Ian McLeod
Abstract:
The merits of the modelling philosophy of Box \& Jenkins (1970) are illustrated with a summary of our recent work on seasonal river flow forecasting. Specifically, this work demonstrates that the principle of parsimony, which has been questioned by several authors recently, is helpful in selecting the best model for forecasting seasonal river flow. Our work also demonstrates the importance of mode…
▽ More
The merits of the modelling philosophy of Box \& Jenkins (1970) are illustrated with a summary of our recent work on seasonal river flow forecasting. Specifically, this work demonstrates that the principle of parsimony, which has been questioned by several authors recently, is helpful in selecting the best model for forecasting seasonal river flow. Our work also demonstrates the importance of model adequacy. An adequate model for seasonal river flow must incorporate seasonal periodic correlation. The usual autoregressive-moving average (ARMA) and seasonal ARMA models are not adequate in this respect for seasonal river flow time series. A new diagnostic check, for detecting periodic correlation in fitted ARMA models is developed in this paper. This diagnostic check is recommended for routine use when fitting seasonal ARMA models. It is shown that this diagnostic check indicates that many seasonal economic time series also exhibit periodic correlation. Since the standard forecasting methods are inadequate on this account, it can be concluded that in many cases, the forecasts produced are sub-optimal. Finally, a limitation of the arbitrary combination of forecasts is also illustrated. Combining forecasts from an adequate parsimonious model with an inadequate model did not improve the forecasts whereas combining the two forecasts of two inadequate models did yield an improvement in forecasting performance. These findings also support the model building philosophy of Box \& Jenkins. The non-intuitive findings of Newbold \& Granger (1974) and Winkler \& Makridakis (1983) that the apparent arbitrary combination of forecasts from similar models will lead to forecasting performance is not supported by our case study with river flow forecasting.
△ Less
Submitted 4 November, 2016;
originally announced November 2016.
-
Necessary and Sufficient Condition for Nonsingular Fisher Information Matrix in ARMA Models
Authors:
A. Ian McLeod
Abstract:
It is demonstrated that a necessary and sufficient condition that the Fisher information matrix of an ARMA model be nonsingular is that the model not be redundant, that is, the autoregressive and moving-average polynomials do not share common roots.
It is demonstrated that a necessary and sufficient condition that the Fisher information matrix of an ARMA model be nonsingular is that the model not be redundant, that is, the autoregressive and moving-average polynomials do not share common roots.
△ Less
Submitted 4 November, 2016;
originally announced November 2016.
-
Faster ARMA maximum likelihood estimation
Authors:
A. Ian McLeod,
Ying Zhang
Abstract:
A new likelihood based AR approximation is given for ARMA models. The usual algorithms for the computation of the likelihood of an ARMA model require $O(n)$ flops per function evaluation. Using our new approximation, an algorithm is developed which requires only $O(1)$ flops in repeated likelihood evaluations. In most cases, the new algorithm gives results identical to or very close to the exact m…
▽ More
A new likelihood based AR approximation is given for ARMA models. The usual algorithms for the computation of the likelihood of an ARMA model require $O(n)$ flops per function evaluation. Using our new approximation, an algorithm is developed which requires only $O(1)$ flops in repeated likelihood evaluations. In most cases, the new algorithm gives results identical to or very close to the exact maximum likelihood estimate (MLE). This algorithm is easily implemented in high level Quantitative Programming Environments (QPEs) such as {\it Mathematica\/}, MatLab and R. In order to obtain reasonable speed, previous ARMA maximum likelihood algorithms are usually implemented in C or some other machine efficient language. With our algorithm it is easy to do maximum likelihood estimation for long time series directly in the QPE of your choice. The new algorithm is extended to obtain the MLE for the mean parameter. Simulation experiments which illustrate the effectiveness of the new algorithm are discussed. {\it Mathematica\/} and R packages which implement the algorithm discussed in this paper are available (McLeod and Zhang, 2007). Based on these package implementations, it is expected that the interested researcher would be able to implement this algorithm in other QPE's.
△ Less
Submitted 3 November, 2016;
originally announced November 2016.
