Search

This paper considers how well the approach of combining forecasts extends to the context of electricity prices. With the increasing popularity of regime switching and time-varying parameter models for predicting power prices, the multi model and evolutionary considerations that usually support the combining of simpler time series methods may be less applicable when the individual models incorporate these features. We address this question with a backtesting analysis on British day-ahead prices. Furthermore, given the volatility of power prices and concerns about accurate forecasting under extreme price excursions, we evaluate the results using various error metrics including expected shortfall. The comparisons are furthermore carefully simulated to consider model selection uncertainty in order to realistically test the value of combining as an ex ante policy. Overall, our results support combining for both accurate operational planning and risk management

This paper focuses on the day-ahead forecasting performance of some models for hourly electricity spot prices. In particular, our rst aim is to investigate the forecasting performance of the model proposed by Huisman et al. (2007) to model hourly power prices in a panel framework. Secondly, we want to extend the model specifying a GARCH structure to take into account a possible conditional heteroskedasticity. A further extension concerns the introduction of cross-lagged correlations into the model through a VAR-type specication. Finally a non linear Markov switching structure is considered to take into account the occurrence of price spikes. We consider price data from di erent european electricity power markets and obtain forecasts for one-day-ahead up to one-week-ahead horizons. For forecast comparisons we use the RMSPE, the MAE and the Diebold and Mariano test. Results show that some improvement can be obtained by GARCH specication in the one-day ahead prediction exercise and by the VAR and MS models in longer prediction horizons.

In this paper we present an extensive comparison of four different classes of models for daily forecasting of spot electricity prices, including ARMAX, constant and time-varying parameter regression models as well as non linear Markov regime-switching regressions. They are selected for particular reasons related to the emerging body of research on the price formation processes observed in electricity markets. The analyses are conducted for representative trading periods of the day in the UK Power Exchange prompt market, with the price series adjusted for their deterministic components and spikes. They show that relative out-of-sample forecasting performances are distinctly different for each trading period, season and across the actual performance metrics. No model consistently outperforms the others, but the ARMAX approach performs well in most cases and the Diebold and Mariano test indicates that, when it is not the best, the ARMAX model is not statistically different from the best. Nevertheless, we suggest that subtle differences in performance between different methods under different conditions are consistent with the apparent variations in the price formation processes by time of day and by season. We conclude with some observations on the disparities between the model specifications appropriate for understanding in-sample price formation and those for accurate out-of-sample predictions.

Goal of this paper is to analyze models to forecast the realized volatility. In particular, we propose a dynamics that simultaneously captures long memory and nonlinearities in which level and persistence shifts through a Markov switching approach. We propose an efficient Markov chain Monte Carlo (MCMC) algorithm to estimate parameters and latent processes involved. Also Bayesian predictive densities have been obtained within such algorithm. The in-sample results show that both long memory and nonlinearities are significant and improve the description of the data. The out-sample results, obtained using several forecast horizons, show that introducing these nonlinearities produces superior long range forecasts over those obtained from nested models.

In this work we investigate an alternative bootstrap approach based on a result of Ramsey (1974) and on the Durbin-Levinson algorithm to obtain surrogate series from linear Gaussian processes with long range dependence. We compare this bootstrap method with other existing procedures in a wide Monte Carlo experiment by estimating, parametrically and semiparametrically, the memory parameter d. We consider Gaussian and non-Gaussian processes to prove the robustness of the method to deviations from Normality. The approach is useful also to estimate condence intervals for the memory parameter d by improving the coverage level of the interval.

In this paper we propose a sequential Monte Carlo algorithm to estimate a stochastic volatility model with leverage effect, non constant conditional mean and jumps. Our idea relies on the auxiliary particle filter algorithm together with the Markov Chain Monte Carlo (MCMC) methodology. Our method allows to sequentially evaluate the parameters and the latent processes involved in the dynamics of interest. An empirical application on simulated data and on the Standard & Poor's 500 index is presented to study the performance of the algorithm implemented.

In this work we introduce a new bootstrap approach based on a result of Ramsey (1974) and on the Durbin-Levinson algorithm to obtain surrogate series from linear Gaussian processes with long range dependence. First we investigate properties of this type of bootstrap, then we apply the method to semi-parametric estimators of the long memory parameter. We find out that the performance of our bootstrap procedure is superior, in terms of MSE, to other established approaches.

In this paper we propose a sequential Monte Carlo algorithm to estimate a stochastic volatility model with leverage effect and non constant conditional mean. Our proposal relies on the auxiliary particle filter method that allows to sequentially evaluate the parameters of the latent processes involved in the dynamics of interest. An empirical application on simulated data is presented to study the performance of the implemented algorithm

A distinguishing feature of the intra-day time-varying volatility of financial time series is given by the presence of long-range dependence of periodic type due mainly to time-of-the-day phenomena. In this work we introduce a model able to describe the empirical evidence given by this periodic longmemory behaviour. The model, named PLM-GARCH (Periodic Long Memory GARCH), represents a natural extension of the FIGARCH model proposed for modelling long-range persistence of the volatility of financial time series. Periodic long memory versions of EGARCH (PLM-EGARCH) models are also considered. Some properties and characteristics of the models are given and an estimation procedure based on quasi maximum likelihood is established. Further possible extensions of the model to take into account multiple sources of periodic long-memory behaviour are suggested. Some empirical applications on intra-day financial time series are also provided.

In this work we introduce an extension of the FIGARCH model, called SFIGARCH, to account for periodic and long memory components in intraday volatility of financial returns. An application on a real time series is provided.

Process capability indices (PCIs) have been widely used in manufacturing industries to provide a quantitative measure of process potential and performance. While some efforts have been dedicated in the literature to the statistical properties od PCIs estimators, scarce attention has been given to the evaluation of these properties when sample data are affected by measurement errors. In this work we deal with the problem of measurement errors effects on the performance of PCIs. The analysis is illustrated with reference to Cp and Cpk, i.e. the two most common measures suggested to evaluate process capability.

One of the main concerns in air pollution is excessive tropospheric ozone concentration. The aim of this work is to develop statistical models giving short-term prediction of future ground-level ozone concentrations. Since there are few physical insights about the dynamic relationship between ozone, precursor emissions and/or meteorological factors, a nonparametric and nonlinear approach sems promising in order to specify the prediction models. First, we apply four nonparametric procedures to forecast daily maximum 1-hour and maximum 8-hours averages of ozone concentrations in an urban area. Then, in order to improve the prediction performances, we combine the time series of the forecasts. This idea seems to give promising results.