我有一个每周销售价值的时间序列对象,并使用KPSS测试和ADF测试测试了平稳性。两项测试都告诉我数据是静止的。
> kpss.test(salests)
KPSS Test for Level Stationarity
data: salests
KPSS Level = 0.34151, Truncation lag parameter = 2, p-value = 0.1
> adf.test(salests)
Augmented Dickey-Fuller Test
data: salests
Dickey-Fuller = -4.9851, Lag order = 4, p-value = 0.01
alternative hypothesis: stationary
但是,当我将时间序列放在auto.arima()
时,它会返回一个d = 1的模型。有谁能解释一下?数据是每周销售,我在xreg
参数中使用多个回归量。
> fit
Series: traints
Regression with ARIMA(0,1,1) errors
Coefficients:
ma1 as.factor(proj$Brand)1 as.factor(proj$Brand)2 Bundle
-0.9438 -167.1745 52.8263 99.4438
s.e. 0.0296 56.7019 48.3095 46.7456
as.factor(proj$Reduction)0.25 as.factor(proj$Reduction)0.33
177.7417 541.3828
s.e. 27.6599 52.7583
sigma^2 estimated as 16570: log likelihood=-669.62
AIC=1353.23 AICc=1354.36 BIC=1371.94