Urgent Please ! different estimates for the same equation

[danongohou]

Can someone please help! I run an eqution (see below) times ago, today i use the same equation with the same times series but i have different estimates.
Even with totaly different equation I have unbelivable estimates. I can see that the number of iteration is different. Please help!
Note that i am using Eviews 6 student version.

original estimates

Date: 01/23/11 Time: 09:07
Sample (adjusted): 2001M01 2009M12
Included observations: 108 after adjustments
Convergence achieved after 4 iterations
White Heteroskedasticity-Consistent Standard Errors & Covariance
BCEAO_RATE=C(2)*BCEAO_RATE(-1)+(1-C(2))*(C(1)+C(3)*INFLATION
+C(4)*OUTPUTGAP_WAEMU(-1)+C(5)*ECB_RATE(-1)+C(6)
*RER_EUROUSD(-1))

CoefficientStd. Error t-Statistic Prob.

C(2) 0.941422 0.041054 22.93130 0.0000
C(1) -0.298432 0.783826 -0.380738 0.7042
C(3) 0.200179 0.141566 1.414034 0.1604
C(4) -0.003374 0.013925 -0.242324 0.8090
C(5) 0.331134 0.136961 2.417729 0.0174
C(6) 1.957204 0.348823 5.610880 0.0000



Second estimation


Dependent Variable: BCEAO_RATE
Method: Least Squares
Date: 02/05/11 Time: 07:59
Sample (adjusted): 2001M01 2009M12
Included observations: 108 after adjustments
Convergence achieved after 1 iteration
White Heteroskedasticity-Consistent Standard Errors & Covariance
BCEAO_RATE=C(2)*BCEAO_RATE(-1)+(1-C(2))*(C(1)+C(3)*INFLATION
+C(4)*OUTPUTGAP_WAEMU(-1)+C(5)*ECB_RATE(-1)+C(6)
*RER_EUROUSD(-1))

CoefficientStd. Error t-Statistic Prob.

C(2) 1.000004 0.043592 22.94001 0.0000
C(1) 10436.65 1.04E+08 0.000101 0.9999
C(3) -2751.872 27305384 -0.000101 0.9999
C(4) -162.3330 1610831. -0.000101 0.9999
C(5) -3170.221 31457873 -0.000101 0.9999
C(6) 6570.202 65170151 0.000101 0.9999

[EViews Gareth]

You're performing non-linear least squares. Your estimates are probably quite sensitive to starting values.

[danongohou]

Thanks a lot, in this case what should i do? I already restarted the computeur...
Regards

[EViews Gareth]

Just make sure you're using the same starting values each time. Starting values are stored in the C vector in your workfile.

[startz]

Your second estimates suggest you have a random walk. When c(2)=1, none of the other coefficients are identified.