R help - Nov 2007 - problem modeling time series

Good afternoon!

I'm trying to model a ts but unfortunately i'.m very new to this kind of
modeling  so i 'll be very grateful if you have an advice.

This was my syntax:

t<-c(16115,17391,19011,20256,19034,18851,20016,18088,19166,21163,18463,19397,15800,16113,18879,20598,17252,19753,19110,19605,20836,18868,20204,24384,15817,18223,19884,21059,18545,19853,20027,20061,21679,20210,20351,21322,16891,17111,20166,18735,16821,17891,17058,19250)
plot.ts(t)
acf(t)
pacf(t)
arima(t,order=c(13,2,0), seasonal=list(order=c(1,1,1)))->fit
predict(fit, n.ahead=1) 


Now, choosing my order, was a trial and error process where i used the acf, pcf
and AIC  (which  is minimized by taking those specific values)  but as you can
imagine   this is something "made by ear". I want to know if there is
some test which can give me the proper values for the order? (i haven't
found some full examples which describe the process fully from head to toe)

Second, if this is the best fitting model for that specific ts, now don't
you think that those standard errors are huge!!!!!!!!!!!!! Say, if the novice
client eliminates  that  great rise/damp in the series and wants to
"predict" solely based on his impressions from the time series he
would probably give me the same interval, as the arima gave me, and i can't
give him some  new  piece of info.  Is there something wrong with my ts, or with
my arima model, and  how  could i make that confidence  interval smaller ?

Thank you and have a great day! 

       
---------------------------------

	[[alternative HTML version deleted]]

You appear to have fitted a seasonal model to a non-seasonal time series.
See ?ts for how to set up a time series (a step you seem to have omitted), 
and ?arima for how to specify the model in R.

Otherwise, this is not the place for a tutorial on fitting time series, 
and many of us do not offer such help to people who do not supply their 
credentials (see the posting guide).  Unless this is a homework problem, a 
time series is more than a series of numbers: it has a time base and a 
context that will influence how it should be modelled.

On Tue, 20 Nov 2007, eugen pircalabelu wrote:
> Good afternoon!
>
> I'm trying to model a ts but unfortunately i'.m very new to this
kind of modeling  so i 'll be very grateful if you have an advice.
>
> This was my syntax:
>
>
t<-c(16115,17391,19011,20256,19034,18851,20016,18088,19166,21163,18463,19397,15800,16113,18879,20598,17252,19753,19110,19605,20836,18868,20204,24384,15817,18223,19884,21059,18545,19853,20027,20061,21679,20210,20351,21322,16891,17111,20166,18735,16821,17891,17058,19250)
> plot.ts(t)
> acf(t)
> pacf(t)
> arima(t,order=c(13,2,0), seasonal=list(order=c(1,1,1)))->fit
> predict(fit, n.ahead=1)
>
>
> Now, choosing my order, was a trial and error process where i used the acf,
pcf and AIC  (which  is minimized by taking those specific values)  but as you
can imagine   this is something "made by ear". I want to know if there
is some test which can give me the proper values for the order? (i haven't
found some full examples which describe the process fully from head to toe)
>
> Second, if this is the best fitting model for that specific ts, now
don't you think that those standard errors are huge!!!!!!!!!!!!! Say, if the
novice client eliminates  that  great rise/damp in the series and wants to
"predict" solely based on his impressions from the time series he
would probably give me the same interval, as the arima gave me, and i can't
give him some  new  piece of info.  Is there something wrong with my ts, or with
my arima model, and  how  could i make that confidence  interval smaller ?
>
> Thank you and have a great day!
>
>
> ---------------------------------
>
> 	[[alternative HTML version deleted]]
>
> ______________________________________________
> R-help at r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/r-help
> PLEASE do read the posting guide
http://www.R-project.org/posting-guide.html
> and provide commented, minimal, self-contained, reproducible code.
>
-- 
Brian D. Ripley,                  ripley at stats.ox.ac.uk
Professor of Applied Statistics,  http://www.stats.ox.ac.uk/~ripley/
University of Oxford,             Tel:  +44 1865 272861 (self)
1 South Parks Road,                     +44 1865 272866 (PA)
Oxford OX1 3TG, UK                Fax:  +44 1865 272595