[R] Incoherence between arima.sim and auto.arima

[Giusy]

Hello,
I have a question about function arima.sim
I tried to somulate a AR(1) process, with no innovation, no error term.
I used this code:

library(forecast)
e=rnorm(100,mean=0,sd=0)
series=arima.sim(model=list(ar=0.75),n=100,innov=e)+20

Then I tried to applicate ti this series auto.arima function:

mod1<-auto.arima(series,stepwise=FALSE,trace=TRUE,ic='aicc')

The best model returned is ARIMA (3,2,1), that is different from (1,0,0)
that I used to generate the series...
Besides, I tried to force a model (1,0,0), and then I compared the fitted. I
attached my plot
The strange thing is that the ARIMA(3,2,1) fitted my series better than
ARIMA (1,0,0) that I used to generate this...
The code is:

mod2<-arima(series,c(1,0,0))
plot(series,main='black=series - red=ARIMA(3,2,1) - green=ARIMA(1,0,0)')
lines(fitted(mod1),col='red')
lines(fitted(mod2),col='green')


Did I do somothing wrong? Anyone can give me explanation?
Thank you very much

http://www.nabble.com/file/p25663200/plot.PNG plot.PNG 
-- 
View this message in context: http://www.nabble.com/Incoherence-between-arima.sim-and-auto.arima-tp25663200p25663200.html
Sent from the R help mailing list archive at Nabble.com.