I cannot describe the procedure step by step because I don't have SPSS presently on my computer. It is certainly possible at least using syntax commands and with some restrictions. I looked into my past output and could find the following commands :
ARIMA VARIABLES= Electricity WITH OutsideTemp1 Work WorkOutsideTemp1
/MODEL =(3,0,0)(0,1,0)96
CONSTANT
/AR=0.31, 0.16, 0.14
/REG= 0.782, 110.0, 4.8
/MXITER=30
/CINPCT=95
/FORECAST=EXACT
EXECUTE.
OutsideTemp1 is the lagged by one period (quarter, I think) outside temperature. So you may need to define lagged variables before using the ARIMA command. In this case it is an AR process but with a seasonal difference. The initial values for the AR and the regression parameters have been set but it may be unnecessary. I remember I had to look in the documentation, in particular the syntax of the ARIMA command. Be careful when writing the equation which correspond to the results. Good luck.
Based on my experience, SPSS doesn't have it. Of course, I have worked with SPSS v19. Minitabe is a better software for time series analysis. It includes ARIMA and SARIMA. However, for ARIMAX, I think you need to use a package in R.
Dear Ramin. I am afraid you are wrong. See my answer above.What can be true is that it is not available within the interface.
As far as I know, the ARIMA command in SPSS was designed by Craig Ansley, an expert in the domainof regression with ARIMA errors. I feel that the more recent command TSMODEL is less powerful and less accurate for some aspects. I have produced a poster comparing several packages (but not Minitab) for the treatment of long time series. Let me know if you are interested.