phase problems

[savvaskef]

Hi, I tried to use forecast in a dataset that is used in a tutorial.so the data have an autoregression at phases 2,3,4,6,12. There is just a small variation so as to avoid problems in confidence intervals.The result is quite intriguing. The fit function produces good results with a phase of 1 and thus seasonality dampened to a line. Its’s such an easy problem that it can pass under the radar I guess.See what you can tell me about it: I enlist the r commands just below.

datset_freq2<-c(3030,1830,3030,1830,3030,1830,3030,1830,3030,1830,3030,1830,3060,1830,3030,1830,3030,1830,3030,1830,3030,1830,3030,1830,3030,1830,3030,1830,3030,1830,3030,1830,3030,1830,3030,1830) timeseriesdataset2 <-ts(datset_freq2,frequency=12) fit<-auto.arima(y=timeseriesdataset2) fcast <- forecast(fit,h=12,level=c(99,95))

plot(fcast)

you can observe the behavior I describe by commparing

fcast$fitted with timeseriesdataset2 thnx for your time Hi, I tried to use forecast in a dataset that is used in a tutorial.so the data have an autoregression at phases 2,3,4,6,12. There is just a small variation so as to avoid problems in confidence intervals.The result is quite intriguing. The fit function produces good results with a phase of 1 and thus forecast is dampened to a line. Its’s such an easy problem that it can pass under the radar I guess.See what you can tell me about it: I enlist the r commands just below.

datset_freq2<-c(3030,1830,3030,1830,3030,1830,3030,1830,3030,1830,3030,1830,3060,1830,3030,1830,3030,1830,3030,1830,3030,1830,3030,1830,3030,1830,3030,1830,3030,1830,3030,1830,3030,1830,3030,1830) timeseriesdataset2 <-ts(datset_freq2,frequency=12) fit<-auto.arima(y=timeseriesdataset2) fcast <- forecast(fit,h=12,level=c(99,95))

plot(fcast)

you can observe the behavior I describe by commparing

fcast$fitted with timeseriesdataset2

thnx for your time

[robjhyndman]

The problem is that auto.arima is not picking up the seasonality, and is fitting an ARIMA(0,1,0) model.