[EMAIL PROTECTED] (Tugba Taskaya) wrote in message news:<[EMAIL PROTECTED]>...
> I recently started to work on "SARIMA MODELS".
> I am trying to reimplement an already implemented experiment which
> was published on a paper called "Combining neural network model with
> seasonal time series ARIMA Model". In the paper, they give the
> equation they used for SARIMA as follows: (based on
> ARIMA(0,1,1)(0,1,0)12, they got)
>
> (1-B)(1-B^12)Z(t) = (1-0.88126B)a(t)
>
> Z(t) denotes the observed value at time t, a(t) is the estimated
> residual at time t.B is the backward shift operator.a(t) should be
> independently distributed as normal random variables with mean=0 and
> variance v^2.
> A part of the data set I am using is (taken by Taiwan machinery
> industry time series):
> t-14 24206.5
> t-13 26075.5
> t-12 21372
> t-1114555
> t-10 19824
> t-9 21115
> t-8 22414
> t-7 21805
> t-6 22172
> t-5 22217
> t-4 21255
> t-3 21333
> t-2 21884
> t-1 25191
> t 17598
> Could you please tell me what is the forecast of SARIMA model at
> time t and the residual a(t) according to the equation and data given?
> Thanks!
I interpret your question to be that you want a forecast for time
period t+1 from ( after observing the reading at time period t )
You have insufficient information to prepare the forecast for time
period t+1 .
The prediction at time period is
ypred(t+1)= y(t)+y(t-11)-y(t-12)-.88126*a(t)
or
= 17598+1114555-21372-.88126*a(t)
you must specify a(t) .
You can do that by iterating the equation at each of the t-1
historical origins starting at time period 14 after setting
a(1),a(2),,,,,a(13) equal to 0.
Also the coefficient of .888126 ( nearly 1.0 ! ) indicates a potential
OVER DIFFERENCING OF ORDER 1 or the presence of a deterministic
component such as a local time trend or a level shift.
Hope this helps ...
Dave Reilly
Automatic Forecasting Systems
* the home of AUTOBOX *
http://www.autobox.com
.
.
=================================================================
Instructions for joining and leaving this list, remarks about the
problem of INAPPROPRIATE MESSAGES, and archives are available at:
. http://jse.stat.ncsu.edu/ .
=================================================================
