我目前正在处理VALE3.SA https://finance.yahoo.com/quote/VALE3.SA/history?p=VALE3.SA的时间序列。
在R上,程序包预测的auto.arima()函数返回ARIMA(2,0,0)。但是当我使用adf.test()和pp.test()时,它们表示不稳定。我还尝试了diff(),这使我的序列变成平稳的,但是当我尝试强制auto.arima使用1个微分时,它表明ARIMA(0,1,0)。我应该使用什么型号?没有平稳性,ARIMA(2,0,0)甚至有效吗?
auto <- auto.arima(tsvale[,"Close"])
ARIMA(2,0,0) with non-zero mean
Coefficients:
ar1 ar2 mean
1.1156 -0.1518 35.9059
s.e. 0.0739 0.0742 6.0635
sigma^2 estimated as 10.89: log likelihood=-464.95
AIC=937.91 AICc=938.14 BIC=950.63 ```
> adf.test(tsvale[,"Close"])
Augmented Dickey-Fuller Test
data: tsvale[, "Close"]
Dickey-Fuller = -2.2874, Lag order = 5, p-value = 0.4561
alternative hypothesis: stationary
> pp.test(tsvale[,"Close"])
Phillips-Perron Unit Root Test
data: tsvale[, "Close"]
Dickey-Fuller Z(alpha) = -8.1416, Truncation lag parameter = 4, p-value = 0.649
alternative hypothesis: stationary```
auto.arima(tsvale[,"Close"], d = 1)
Series: tsvale[, "Close"]
ARIMA(0,1,0)
sigma^2 estimated as 11.13: log likelihood=-464.39
AIC=930.78 AICc=930.8 BIC=933.96 ```