Forecasting with serially correlated regression models

In this article we investigate the asymptotic and finite-sample properties of predictors of regression models with autocorrelated errors. We prove new theorems associated with the predictive efficiency of generalized least squares (GLS) and incorrectly structured GLS predictors. We also establish the form associated with their predictive mean squared errors as well as the magnitude of these errors relative to each other and to those generated from the ordinary least squares (OLS) predictor. A large simulation study is used to evaluate the finite-sample performance of forecasts generated from models using different corrections for the serial correlation.

A simple regional-scale model for forecasting sorghum yield across North-Eastern Australia

Sorghum is the main dryland summer crop in NE Australia and a number of agricultural businesses would benefit from an ability to forecast production likelihood at regional scale. In this study we sought to develop a simple agro-climatic modelling approach for predicting shire (statistical local area) sorghum yield. Actual shire yield data, available for the period 1983-1997 from the Australian Bureau of Statistics, were used to train the model. Shire yield was related to a water stress index (SI) that was derived from the agro-climatic model. The model involved a simple fallow and crop water balance that was driven by climate data available at recording stations within each shire. Parameters defining the soil water holding capacity, maximum number of sowings (MXNS) in any year, planting rainfall requirement, and critical period for stress during the crop cycle were optimised as part of the model fitting procedure. Cross-validated correlations (CVR) ranged from 0.5 to 0.9 at shire scale. When aggregated to regional and national scales, 78-84% of the annual variation in sorghum yield was explained. The model was used to examine trends in sorghum productivity and the approach to using it in an operational forecasting system was outlined. (c) 2005 Elsevier B.V. All rights reserved.

Parameter setting and statistical learning

Three main models of parameter setting have been proposed: the Variational model proposed by Yang (2002; 2004), the Structured Acquisition model endorsed by Baker (2001; 2005), and the Very Early Parameter Setting (VEPS) model advanced by Wexler (1998). The VEPS model contends that parameters are set early. The Variational model supposes that children employ statistical learning mechanisms to decide among competing parameter values, so this model anticipates delays in parameter setting when critical input is sparse, and gradual setting of parameters. On the Structured Acquisition model, delays occur because parameters form a hierarchy, with higher-level parameters set before lower-level parameters. Assuming that children freely choose the initial value, children sometimes will miss-set parameters. However when that happens, the input is expected to trigger a precipitous rise in one parameter value and a corresponding decline in the other value. We will point to the kind of child language data that is needed in order to adjudicate among these competing models.

Mortality variation across Australia: descriptive data for States and Territories, and statistical divisions

OBJECTIVE: To describe variation in all cause and selected cause-specific mortality rates across Australia. METHODS: Mortality and population data for 1997 were obtained from the Australian Bureau of Statistics. All cause and selected cause-specific mortality rates were calculated and directly standardised to the 1997 Australian population in 5-year age groups. Selected major causes of death included cancer, coronary artery disease, cerebrovascular disease, diabetes, accidents and suicide. Rates are reported by statistical division, and State and Territory. RESULTS: All cause age-standardised mortality was 6.98 per 1000 in 1997 and this varied 2-fold from a low in the statistical division of Pilbara, Western Australia (5.78, 95% confidence interval 5.06-6.56), to a high in Northern Territory-excluding Darwin (11.30, 10.67-11.98). Similar mortality variation (all p<0.0001) exists for cancer (1.01-2.23 per 1000) and coronary artery disease (0.99-2.23 per 1000), the two biggest killers. Larger variation (all p<0.0001) exists for cerebrovascular disease (0.7-11.8 per 10,000), diabetes (0.7-6.9 per 10,000), accidents (1.7-7.2 per 10,000) and suicide (0.6-3.8 per 10,000). Less marked variation was observed when analysed by State and Territory. but Northern Territory consistently has the highest age-standardised mortality rates. CONCLUSIONS: Analysed by statistical division, substantial mortality gradients exist across Australia, suggesting an inequitable distribution of the determinants of health. Further research is required to better understand this heterogeneity.

An object-oriented neural network approach to short-term traffic forecasting

This paper discusses an object-oriented neural network model that was developed for predicting short-term traffic conditions on a section of the Pacific Highway between Brisbane and the Gold Coast in Queensland, Australia. The feasibility of this approach is demonstrated through a time-lag recurrent network (TLRN) which was developed for predicting speed data up to 15 minutes into the future. The results obtained indicate that the TLRN is capable of predicting speed up to 5 minutes into the future with a high degree of accuracy (90-94%). Similar models, which were developed for predicting freeway travel times on the same facility, were successful in predicting travel times up to 15 minutes into the future with a similar degree of accuracy (93-95%). These results represent substantial improvements on conventional model performance and clearly demonstrate the feasibility of using the object-oriented approach for short-term traffic prediction. (C) 2001 Elsevier Science B.V. All rights reserved.

Statistical questions - Comment

No Abstract

Forecasting regional crop production using SOI phases: an example for the Australian peanut industry

Using peanuts as an example, a generic methodology is presented to forward-estimate regional crop production and associated climatic risks based on phases of the Southern Oscillation Index (SOI). Yield fluctuations caused by a highly variable rainfall environment are of concern to peanut processing and marketing bodies. The industry could profitably use forecasts of likely production to adjust their operations strategically. Significant, physically based lag-relationships exist between an index of ocean/atmosphere El Nino/Southern Oscillation phenomenon and future rainfall in Australia and elsewhere. Combining knowledge of SOI phases in November and December with output from a dynamic simulation model allows the derivation of yield probability distributions based on historic rainfall data. This information is available shortly after planting a crop and at least 3-5 months prior to harvest. The study shows that in years when the November-December SOI phase is positive there is an 80% chance of exceeding average district yields. Conversely, in years when the November-December SOI phase is either negative or rapidly falling there is only a 5% chance of exceeding average district yields, but a 95% chance of below average yields. This information allows the industry to adjust strategically for the expected volume of production. The study shows that simulation models can enhance SOI signals contained in rainfall distributions by discriminating between useful and damaging rainfall events. The methodology can be applied to other industries and regions.

An adaptive neural-wavelet model for short term load forecasting

This paper proposed a novel model for short term load forecast in the competitive electricity market. The prior electricity demand data are treated as time series. The forecast model is based on wavelet multi-resolution decomposition by autocorrelation shell representation and neural networks (multilayer perceptrons, or MLPs) modeling of wavelet coefficients. To minimize the influence of noisy low level coefficients, we applied the practical Bayesian method Automatic Relevance Determination (ARD) model to choose the size of MLPs, which are then trained to provide forecasts. The individual wavelet domain forecasts are recombined to form the accurate overall forecast. The proposed method is tested using Queensland electricity demand data from the Australian National Electricity Market. (C) 2001 Elsevier Science B.V. All rights reserved.