TESTING THE SIGNIFICANCE OF CORRELATION BETWEEN TIME SERIES
成果类型：
Article
署名作者：
ORCUTT, GH; JAMES, SF
刊物名称：
BIOMETRIKA
ISSN/ISSBN：
0006-3444
DOI：
10.1093/biomet/35.3-4.397
发表日期：
1948
页码：
397413
关键词：
摘要：
Prior investigation of 52 economic time series indicated they might have been drawn from a population of series generated by [PSI]t+1=[PSI] t + .3([PSI]t- [PSI]t-1) + [epsilon]t+1, where [epsilon] is random in time with expected value of zero. Using this model, a set of non-related series were constructed and correlations between them were computed. Empirical investigation indicates that significance tests may be devised by computing the approx. variance of the correlation r between 2 unrelated time series in terms of their first lag auto-correlations as var [image] and referring r to the distribution of the product moment correlation in samples of m from an uncorrelated universe, where m is chosen so that the variance of the product moment coeff. is approximately equal to the calculated variance. In the example considered, in spite of the fact that nonstationary series having small n and not generated by a Markoff process were used, the approximations are very good.
来源URL：