A Study of the Autoregressive Nature of the Time Series Used for Tinbergen's Model of the Economic System of the United States, 1919-1932
A Study of the Autoregressive Nature of the Time Series Used for Tinbergen's Model of the Economic System of the United States, 1919-1932
复制标题
用于 1919-1932 年美国经济体系廷伯根模型的时间序列的自回归性质研究
DOI：
10.1111/j.2517-6161.1948.tb00001.x
复制
发表时间：
1948
期刊：
影响因子：
--
通讯作者：
G. Orcutt
中科院分区：
文献类型：
--
作者：
G. Orcutt
为了进行短期预测、设计显着性检验和开发经济系统的动态模型，更充分地了解经济时间序列的当前值和过去值之间存在的关系将非常有用。因此，给定直到时间 t 的序列值，我们想知道关于该序列的预期未来值，什么样的推论是合理的。无法获得更多此类知识的原因可能有很多，但也许最重要的是，除了极少数例外，可用的时间序列非常短，以至于除了非常简单的关系之外，从单个序列中检测任何内容似乎都没有希望。我们的程序旨在通过处理大量系列来减少这种困难的影响，从而使用整个数字来获得对该系列父群体的一些特征的更准确估计。这是否是一个合理的程序取决于是否有任何理由假设所选系列实际上属于接近单一总体的事物。我们试图通过对线性差分方程组的简短考虑来至少使这种可能性变得合理，然后通过将与我们的经济序列相关的某些分布与与从各种序列总体中抽样获得的序列相关的可比分布进行比较来强化我们的希望。通过这种方式，我们表明我们的真实序列集可能来自单个序列总体，并且我们还得出了该总体参数的估计。
For the purposes of making short run predictions, designing of tests of significance and development of dynamic models of economic systems it would be extremely useful to have a more adequate knowledge of the relation or relations existing between present and past values of economic time series. Thus, given the values of a series up to time t, we should like to know what kind of inferences are reasonable concerning expected future values of the series. There are probably many reasons why more knowledge of this sort is not available, but perhaps the most important is that, with rare exceptions, the time series available are so short that detection of anything from single series, except very simple relationships, does not seem promising. Our procedure is aimed at reducing the effect of this difficulty by treating a large number of series, and thus using the whole number to obtain a more accurate estimate of some of the characteristics of this parent population of series. Whether this is a reasonable procedure or not depends on whether there is any reason to suppose that the chosen series do in fact belong to something approaching a single population. The possibility of this we try to make at least plausible by a short consideration of sets of linear difference equations, and then we reinforce our hopes by a comparison of certain distributions connected with our economic series with comparable distributions connected with series obtained by sampling from a variety of populations of series. In this manner we show that our real set of series could have come from a single population of series, and we also arrive at estimates of the parameters of this population.