Autocorrelation can be either +ve or –ve. One of the popular way to remove autocorrelation is to use ARIMA. By setting variables AR and MA in ARIMA, we can deal with autocorrelation. If autocorrelation of regression model is small, then 1st order lag is created by setting AR =1 or MA =1 and if autocorrelation is high, then 2nd or some higher degree lag is created. Negative autocorrelation might arise because of over-differenced variables. Please note that AR =1 adds a lag dependent variable to the forecasting equation and MA =1 adds a lag forecast error.
自相关可以+ VE或–VE。一个删除相关的流行的方式是使用ARIMA。通过设置变量AR和MA模型，我们可以处理自相关。如果回归模型的自相关是小的，然后第一阶滞后是通过设置AR = 1 = 1，如果自相关或马高了，然后第二或更高程度的滞后是造成。负自相关可能因为过差分变量。请注意，AR = 1添加了一个滞后因变量的预测方程和马= 1添加了一个滞后的预测误差。