Research into Spurious Stochastics in Panels, Generalized Forecast Theory and Evaluation of Models

这是一个基于成就的更新项目。这个项目的目标是为实证模型或经济理论的评估带来一些新的想法；虽然将使用两个已建立的构建模块。第一个来自面板分析，并讨论了动态面板模型规范中出现“虚假随机”的可能性，使用的数据来自规模差异很大的地区，但规模效应没有正确建模。尺寸本身可能是缓慢变化的，因此无法通过使用固定效果来捕捉。最初的模拟表明，这是一个潜在的重要问题。随之而来的一个问题是如何评估面板模型，从而如何比较模型。第二个块是考虑一般的预测理论，具有一般的成本t函数，并强调预测分布函数。出现了预测误差的新测试，并根据自己的目标和知识更好地了解适当的建模策略。目前大多数可用的评估方法，如拟合优度、规格正确性和包络性的度量，都是统计度量。有人会说，更好的方法是考虑新模型或理论对经济的价值，或者至少是它对决策者的帮助。这种方法已经在金融领域被采用，它可以更广泛地用于预测。本文将探讨和分析这一观点的含义

This is an accomplishment based renewal project. The goal of this project is to bring some fresh thoughts on the evaluation of empirical models or to economic theories; although two established building blocks will be used. The first comes from panel analysis and discusses the possibility of `spurious stochastics` occurring in a dynamic panel model specification, using data from regions that differ greatly in size, but where the size effect is not properly modeled. Size itself may be slowly evolving and so not be captured by the use of fixed effects. Initial simulations suggest that this is potentially an important problem. One that follows from it is how to evaluate panel models, and thus how to compare models. The second block is consideration of a general theory of forecasting, with general cost t functions and an emphasis on predictive distribution functions. New tests for forecast errors arise and a better appreciation of the appropriate modeling strategy to use depending on ones objective and knowledge. The majority of the currently available methods of evaluation, such as measures of goodness of fit, correctness of specification and encompassing, are statistical measures. It will be argued that a better approach considers the value to the economy of a new model or theory, or at least the aid it gives to decision makers. This is the approach is already taken in finance and it could be more widely used in forecasting. The implications of this viewpoint will be explored and analyzed