I use arima(1,1,0) to fit prob1, the result shows as follows:
> fitprod1
Call:
arima(x = prod1, order = c(1, 1, 0))
Coefficients:
ar1
-0.0798
s.e. 0.0212
sigma^2 estimated as 1.468: log likelihood = -3561.4, aic = 7126.8
However, I use ARIMA(1,0,0) to fit diff(prod1). The result shows as follows:
> fitprod1=arima(diff(prod1),order = c(1,0,0))
> summary(fitprod1)
Call:
arima(x = diff(prod1), order = c(1, 0, 0))
Coefficients:
ar1 intercept
-0.0806 0.0323
s.e. 0.0212 0.0238
sigma^2 estimated as 1.466: log likelihood = -3560.48, aic = 7126.96
Training set error measures:
ME RMSE MAE MPE MAPE MASE ACF1
Training set 3.60474e-05 1.210953 0.8383019 NaN Inf 0.6345769 -0.0003768319
I want to know:
- why the AR coefficient of this two model are different?
- why ARIMA(1,1,0) doesn't have the intercept? How to write the ARIMA(1,1,0) equation?