Serially Correlated Differences in the Paired Comparison of Time Series
研究了两条时间序列的配对差异服从平稳马尔可夫过程时，序列相关对F检验的影响，发现正序列相关使F分布上尾变厚，负序列相关则变薄。
If observations are taken at the same set of time points on each of two time series whose nonnegative trends are proportional, the F statistic based on the differences of the two series at the time points may be used to test equality of trends against inequality of trends provided the paired differences are independently and identically distributed normal random variables. The present paper investigates the effect on the F test when the paired differences follow a stationary Markov process with normal residuals. It is shown that for positive serial correlation the approximate large-sample distribution of F has a thicker upper tail than when the serial correlation is zero, while for negative serial correlation the upper tail is thinner.