Search results

After briefly reviewing the past history of Bayesian econometrics and Alan Greenspan's (2004) recent description of his use of Bayesian methods in managing policy-making risk, some of the issues and needs that he mentions are discussed and linked to past and present Bayesian econometric research. Then a review of some recent Bayesian econometric research and needs is presented. Finally, some thoughts are presented that relate to the future of Bayesian econometrics.

Several lessons learnt from a Bayesian analysis of basic macroeconomic time-series models are presented for the situation where some model parameters have substantial posterior probability near the boundary of the parameter region. This feature refers to near-instability within dynamic models, to forecasting with near-random walk models and to clustering of several economic series in a small number of groups within a data panel. Two canonical models are used: a linear regression model with autocorrelation and a simple variance components model. Several well-known time-series models like unit root and error correction models and further state space and panel data models are shown to be simple generalizations of these two canonical models for the purpose of posterior inference. A Bayesian model averaging procedure is presented in order to deal with models with substantial probability both near and at the boundary of the parameter region. Analytical, graphical, and empirical results using U.S. macroeconomic data, in particular on GDP growth, are presented.

Jeliazkov and Poirier (2008) analyze the daily incidence of violence during the Second Intifada in a statistical way using an analytical Bayesian implementation of a second-order discrete Markov process. We tackle the same data and modeling problem from our perspective as cognitive scientists. First, we propose a psychological model of violence, based on a latent psychological construct we call “build up” that controls the retaliatory and repetitive violent behavior by both sides in the conflict. Build up is based on a social memory of recent violence and generates the probability and intensity of current violence. Our psychological model is implemented as a generative probabilistic graphical model, which allows for fully Bayesian inference using computational methods. We show that our model is both descriptively adequate, based on posterior predictive checks, and has good predictive performance. We then present a series of results that show how inferences based on the model can provide insight into the nature of the conflict. These inferences consider the base rates of violence in different periods of the Second Intifada, the nature of the social memory for recent violence, and the way repetitive versus retaliatory violent behavior affects each side in the conflict. Finally, we discuss possible extensions of our model and draw conclusions about the potential theoretical and methodological advantages of treating societal conflict as a cognitive modeling problem.

We put forward the idea that for model selection the intrinsic priors are becoming a center of a cluster of a dominant group of methodologies for objective Bayesian Model Selection.

The intrinsic method and its applications have been developed in the last two decades, and has stimulated closely related methods. The intrinsic methodology can be thought of as the long searched approach for objective Bayesian model selection and hypothesis testing.

In this paper we review the foundations of the intrinsic priors, their general properties, and some of their applications.

Vector autoregressions (VAR) combined with Minnesota-type priors are widely used for macroeconomic forecasting. The fact that strong but sensible priors can substantially improve forecast performance implies VAR forecasts are sensitive to prior hyperparameters. But the nature of this sensitivity is seldom investigated. We develop a general method based on Automatic Differentiation to systematically compute the sensitivities of forecasts – both points and intervals – with respect to any prior hyperparameters. In a forecasting exercise using US data, we find that forecasts are relatively sensitive to the strength of shrinkage for the VAR coefficients, but they are not much affected by the prior mean of the error covariance matrix or the strength of shrinkage for the intercepts.

Purpose – This chapter provides details of research that attempts to relate traffic operational conditions on uninterrupted flow facilities (e.g., freeways and expressways) with real-time crash likelihood. Unlike incident detection, the purpose of this line of work is to proactively assess crash likelihood and potentially reduce the likelihood through proactive traffic management techniques, including variable speed limit and ramp metering among others.

Methodology – The chapter distinguishes between the traditional aggregate crash frequency-based approach to safety evaluation and the approach needed for real-time crash risk estimation. Key references from the literature are summarised in terms of the reported effect of different traffic characteristics that can be derived in near real-time, including average speed, temporal variation in speed, volume and lane-occupancy, on crash occurrence.

Findings – Traffic and weather parameters are among the real-time crash-contributing factors. Among the most significant traffic parameters is speed particularly in the form of coefficient of variation of speed.

Research implications – In the traffic safety field, traditional data sources are infrastructure-based traffic detection systems. In the future, if automatic traffic detection systems could provide reliable data at the vehicle level, new variables such as headway could be introduced. Transferability of real-time crash prediction models is also of interest. Also, the potential effects of different management strategies to reduce real-time crash risk could be evaluated in a simulation environment.

Practical implications – This line of research has been at the forefront of bringing data mining and other machine-learning techniques into the traffic management arena. We expect these analysis techniques to play a more important role in real-time traffic management, not just for safety evaluation but also for congestion pricing and alternate routing.

Constructing and evaluating behavioral science models is a complex process. Decisions must be made about which variables to include, which variables are related to each other, the functional forms of the relationships, and so on. The last 10 years have seen a substantial extension of the range of statistical tools available for use in the construction process. The progress in tool development has been accompanied by the publication of handbooks that introduce the methods in general terms (Arminger et al., 1995; Tinsley & Brown, 2000a). Each chapter in these handbooks cites a wide range of books and articles on specific analysis topics.