Model uncertainty and the forecast accuracy of ARMA models: A survey (original) (raw)
Related papers
Bayesian Estimate of Parameters for ARMA Model Forecasting
Tatra Mountains Mathematical Publications, 2020
This paper presents a Bayesian approach to finding the Bayes estimator of parameters for ARMA model forecasting under normal-gamma prior assumption with a quadratic loss function in mathematical expression. Obtaining the conditional posterior predictive density is based on the normal-gamma prior and the conditional predictive density, whereas its marginal conditional posterior predictive density is obtained using the conditional posterior predictive density. Furthermore, the Bayes estimator of parameters is derived from the marginal conditional posterior predictive density.
ARCH models for multi-period forecast uncertainty-a reality check using a panel of density forecasts
Advances in Econometrics
We develop a theoretical framework to compare forecast uncertainty estimated from time series models to those available from survey density forecasts. The sum of the average variance of individual densities and the disagreement, which is the same as the variance of the aggregate density, is shown to approximate the predictive uncertainty from well specified time series models when the variance of the aggregate shocks is relatively small compared to that of the idiosyncratic shocks. We argue that due to grouping error problems, compositional effects of the panel, and other complications, the uncertainty measure has to be estimated from individual densities. Despite numerous reservations on the credibility of time series based measures of forecast uncertainty, we found that during normal times the uncertainty estimates based on ARCH models simulate the subjective survey measure remarkably well. However, during times of regime change and structural break, the two estimates do not overl...
Mind your Ps and Qs! Improving ARMA forecasts with RBC priors
2005
We use an RBC model to help inform ARMA parameters for univariate out-of-sample forecasting of US output. Bayesian estimation methods yield an interpretable structural model with forecasting performance at least competitive with ARMA models estimated with standard techniques.
Measuring Forecast Performance of ARMA and ARFIMA Models.pdf
The classical approach to modelling economic series is to apply the Box -Jenkins approach of ARMA or ARIMA depending on whether the series is stationary or nonstationary. If such series exhibits long memory property, forecast values based on ARIMA model may not be reliable. This studies therefore focused on measuring forecast performance of ARIMA(p,d,q) and ARFIMA (p,d,q) models for stationary type series that exhibit Long memory properties. The UK Pound/US Dollar exchange rate data were analysed by OX 5.1 package using the Root mean Square forecast Error (RMSFE) and Mean Absolute Percentage Forecast Error (MAPFE) as measurement criteria.
Determinants of Multi-period Forecast Uncertainty Using a Panel of Density Forecasts
This paper examines the determinants of inflation forecast uncertainty using a panel of density forecasts from the Survey of Professional Forecasters (SPF). We show that previous studies based on aggregate time series data are biased due to heterogeneity in individual forecasts. Based on a dynamic heterogeneous panel data model, we find that the persistence in forecast uncertainty is much less than what the aggregate time series data would suggest. In addition, the strong link between past forecast errors and the current forecast uncertainty, as often is found in the ARCH literature, is largely lost in the multi-period context with varying forecast horizons. Forecasters are found to pay more attention to recent "news" about inflation than the outdated past forecast errors. We propose a novel way of estimating uncertainty of "news" using Kullback-Leibler Information, and show that it is an important determinant of the current inflation forecast uncertainty. Our results also support Friedman (1977)'s conjecture that higher inflation rate leads to higher inflation uncertainty.
A Kullback-Leibler divergence based comparison of approximate Bayesian estimations of ARMA models
Communications for Statistical Applications and Methods
Autoregressive moving average (ARMA) models involve nonlinearity in the model coefficients because of unobserved lagged errors, which complicates the likelihood function and makes the posterior density analytically intractable. In order to overcome this problem of posterior analysis, some approximation methods have been proposed in literature. In this paper we first review the main analytic approximations proposed to approximate the posterior density of ARMA models to be analytically tractable, which include Newbold, Zellner-Reynolds, and Broemeling-Shaarawy approximations. We then use the Kullback-Leibler divergence to study the relation between these three analytic approximations and to measure the distance between their derived approximate posteriors for ARMA models. In addition, we evaluate the impact of the approximate posteriors distance in Bayesian estimates of mean and precision of the model coefficients by generating a large number of Monte Carlo simulations from the approximate posteriors. Simulation study results show that the approximate posteriors of Newbold and Zellner-Reynolds are very close to each other, and their estimates have higher precision compared to those of Broemeling-Shaarawy approximation. Same results are obtained from the application to real-world time series datasets.
Partially adaptive estimation of ARMA time series models
International Journal of Forecasting, 1989
This paper presents a new partially adaptive estimator of ARMA models which includes least absolute deviation (LAD or L~), least squares, Lp, and 'optimal' Lp as special or limiting cases. This estimator is based upon a generalized t(GT) distribution for the residuals. The GT distribution has a mixture interpretation based on an innovative outlier framework, and results in an estimator with a bounded @function for finite 'degrees of freedom.' Joint estimation of distributional parameters and ARMA parameters allows the estimation technique to 'adjust' or 'adapt' to the data type and provides a natural way to test for normality of residuals. Monte Carlo simulations compare the performance of these estimators across different distributional assumptions including a 'contaminated' normal. The partially adaptive estimators perform well across diverse data types including the innovative outlier and 'contaminated normal' models. The distributions of corresponding Q statistics and likelihood ratio statistics are considered.
Economic forecasting: editors’ introduction
Empirical Economics
The 2-day workshop, organized by Robert M. Kunst and Martin Wagner, drew much more attention than originally expected. This reflects the increased-respectively regained-importance of forecasting not only in practical terms but also as a research topic in the underlying scientific disciplines. Much of this growing interest may be also rooted in increased importance of forecasting in fields such as management science, marketing or supply chain management and may well be driven by methodological developments rooted in several disciplines that could be summarized under labels such as big data, machine learning and the like; the workshop itself had a narrower focus on macroeconomic forecasting. Even with the specific focus on macroeconomics, the papers span a wide portfolio of approaches and applications, ranging from statistical theory to data-driven research. As indicated in the beginning, some of the contributions in this special volume are not related to the workshop, as submission of manuscripts was open.
Selection between models through multi-step-ahead forecasting
Journal of Statistical Planning and Inference, 2010
We present and show applications of two new test statistics for deciding if one ARIMA model provides significantly better h -step-ahead forecasts than another, as measured by the difference of approximations to their asymptotic mean square forecast errors. The two statistics differ in the variance estimate whose square root is the statistic's denominator. Both variance estimates are consistent even when the ARMA components of the models considered are incorrect. Our principal statistic's variance estimate accounts for parameter estimation. Our simpler statistic's variance estimate treats parameters as fixed. The broad consistency properties of these estimates yield improvements to what are known as tests of type. These are tests whose variance estimates treat parameters as fixed and are generally not consistent in our context.
Financial Theory and Practice, 2011
In this paper we present a framework for incorporating uncertainties into economic forecasts which includes the calculation, presentation and evaluation of density forecasts. Using the VECM proposed by Rukelj (2010) as the benchmark model, this framework is applied to forecasting economic activity in Croatia. Future and parameter uncertainties using stochastic simulations are considered in density forecasting with parametric and non-parametric approach in generating random errors. The resulting density forecasts are presented using fan charts, and evaluated by the Kolmogorov-Smirnov test of probability integral transforms of density forecasts. The main …ndings are: parametric and non-parametric approach yield similar results, incorporating parameter uncertainty results in a much wider probability bands of the forecasts and evaluation of density forecasts indicate better performance when only future, without parameter, uncertainties are considered.