Hi send me the data in ascii form ie as a text file. when I see you using z, I think you are using Shumway's method of expression which is the reverse of Brockwell's, ie, the plus and minus signs are reversed. But your constant estimated value is a worry.. rjfh ----- Original Message ----- From: Sanzad Siddique To: gretl-users(a)lists.wfu.edu Sent: Wednesday, June 20, 2012 7:47 PM Subject: [Gretl-users] Wrong Coefficient Estimation for ARIMA Hi, I have tried the ARIMA command in GRETL with the attached time series data. The time series is constructed with the below formula ( a- is normal random values): z(t) = 0.50 + 0.50*z(t-1) + 0.40*z(t-2) + a(t)-0.25*a(t-1) -0.75*a(t-2); with the arima command (arima 2 0 2; 1), for order (2,2) , I am getting the below output: coefficient std. error z p-value --------------------------------------------------------- const 4.96454 0.0275435 180.2 0.0000 *** phi_1 ?0.447238 0.789760 ?0.5663 0.5712 phi_2 ?0.0272188 0.390533 ?0.06970 0.9444 theta_1 0.751567 0.787981 0.9538 0.3402 theta_2 ?0.00643631 0.629626 ?0.01022 0.9918 The output is not showing the correct coefficient values for model parameters. I have attached the input data. Can anyone check the attached dataset and let me know what could be the problem. Thanks, ------------------------------------------------------------------------------ _______________________________________________ Gretl-users mailing list Gretl-users(a)lists.wfu.edu http://lists.wfu.edu/mailman/listinfo/gretl-users

On Wed, 20 Jun 2012, Sanzad Siddique wrote: I'm seeing those results using your data, but I'm not seeing anything similar if I construct data on this basis in gretl. <hansl> nulldata 2000 setobs 7 2006/06/02 series a = normal() series z = 5 # the unconditional mean z = 0.5 + 0.5*z(-1) + 0.40*z(-2) + a - 0.25*a(-1) - 0.75*a(-2) arima 2 0 2 ; z arima 2 0 2 ; z --conditional </hansl> Typical result: const 5.00054 phi_1 0.533660 phi_2 0.357096 theta_1 -0.288852 theta_2 -0.711148 The constant is not close to 0.5, but that depends on how mean terms are handled. See section 23.2 of the Gretl User's Guide. In this case z* = 0.5 + 0.5z* + 0.4z*, so z* = 5. If you want to see the constant given as approx 0.5, use conditional ML. Allin Cottrell