Box-Jenkins rats 8.0 and 9.1
Posted: Sun Dec 11, 2016 6:00 am
Hello,
I don't understand why but with two diferent version of rats, I haven't the same results.
In Rats 8.0:
boxjenk(constant,ar=1, ma=1,maxl,diffs=1,regressor) pib 1960:01 * resexo
# trend trend2
Box-Jenkins - Estimation by ML Gauss-Newton
Convergence in 20 Iterations. Final criterion was 0.0000017 <= 0.0000100
Dependent Variable PIB
Annual Data From 1960:01 To 2014:01
Usable Observations 54
Degrees of Freedom 49
Skipped/Missing (from 55) 1
Centered R^2 0.9995619
R-Bar^2 0.9995261
Uncentered R^2 0.9999962
Mean of Dependent Variable 9.7166476565
Std Error of Dependent Variable 0.9181274693
Standard Error of Estimate 0.0199869509
Sum of Squared Residuals 0.0195744320
Log Likelihood 135.9647
Durbin-Watson Statistic 1.8798
Q(13-2) 9.5167
Significance Level of Q 0.5743072
Variable Coeff Std Error T-Stat Signif
************************************************************************************
1. CONSTANT 7.819026571 0.073574757 106.27322 0.00000000
2. AR{1} 0.917717716 0.055341065 16.58294 0.00000000
3. MA{1} 0.397724272 0.135593868 2.93320 0.00509005
4. TREND 0.087847768 0.005618158 15.63640 0.00000000
5. TREND2 -0.000569857 0.000094838 -6.00872 0.00000023
And in Rats 9.1:
boxjenk(constant,ar=1, ma=1,maxl,diffs=1,regressors) pib 1960:01 * resexo
# trend trend2
Box-Jenkins - Estimation by ML Gauss-Newton
NO CONVERGENCE IN 69 ITERATIONS
LAST CRITERION WAS 0.0000000
SUBITERATIONS LIMIT EXCEEDED.
ESTIMATION POSSIBLY HAS STALLED OR MACHINE ROUNDOFF IS MAKING FURTHER PROGRESS DIFFICULT
TRY HIGHER SUBITERATIONS LIMIT, TIGHTER CVCRIT, DIFFERENT SETTING FOR EXACTLINE OR ALPHA ON NLPAR
RESTARTING ESTIMATION FROM LAST ESTIMATES OR DIFFERENT INITIAL GUESSES MIGHT ALSO WORK
Dependent Variable PIB
Annual Data From 1960:01 To 2014:01
Usable Observations 55
Degrees of Freedom 50
Centered R^2 0.9984907
R-Bar^2 0.9983699
Uncentered R^2 0.9999862
Mean of Dependent Variable 9.6855946343
Std Error of Dependent Variable 0.9382876296
Standard Error of Estimate 0.0378827609
Sum of Squared Residuals 0.0717551788
Log Likelihood 104.6087
Durbin-Watson Statistic 0.5631
Q(13-2) 76.6459
Significance Level of Q 0.0000000
Variable Coeff Std Error T-Stat Signif
************************************************************************************
1. CONSTANT 7.784659094 0.030028584 259.24163 0.00000000
2. AR{1} -0.754375420 0.035138862 -21.46841 0.00000000
3. MA{1} 0.324101626 0.043120686 7.51615 0.00000000
4. TREND 0.090274713 0.002567841 35.15588 0.00000000
5. TREND2 -0.000598410 0.000045431 -13.17190 0.00000000
I have check many times, I used exaclty the same data. But usable Observations and degrees of fredom are diferents.
I don't understand why but with two diferent version of rats, I haven't the same results.
In Rats 8.0:
boxjenk(constant,ar=1, ma=1,maxl,diffs=1,regressor) pib 1960:01 * resexo
# trend trend2
Box-Jenkins - Estimation by ML Gauss-Newton
Convergence in 20 Iterations. Final criterion was 0.0000017 <= 0.0000100
Dependent Variable PIB
Annual Data From 1960:01 To 2014:01
Usable Observations 54
Degrees of Freedom 49
Skipped/Missing (from 55) 1
Centered R^2 0.9995619
R-Bar^2 0.9995261
Uncentered R^2 0.9999962
Mean of Dependent Variable 9.7166476565
Std Error of Dependent Variable 0.9181274693
Standard Error of Estimate 0.0199869509
Sum of Squared Residuals 0.0195744320
Log Likelihood 135.9647
Durbin-Watson Statistic 1.8798
Q(13-2) 9.5167
Significance Level of Q 0.5743072
Variable Coeff Std Error T-Stat Signif
************************************************************************************
1. CONSTANT 7.819026571 0.073574757 106.27322 0.00000000
2. AR{1} 0.917717716 0.055341065 16.58294 0.00000000
3. MA{1} 0.397724272 0.135593868 2.93320 0.00509005
4. TREND 0.087847768 0.005618158 15.63640 0.00000000
5. TREND2 -0.000569857 0.000094838 -6.00872 0.00000023
And in Rats 9.1:
boxjenk(constant,ar=1, ma=1,maxl,diffs=1,regressors) pib 1960:01 * resexo
# trend trend2
Box-Jenkins - Estimation by ML Gauss-Newton
NO CONVERGENCE IN 69 ITERATIONS
LAST CRITERION WAS 0.0000000
SUBITERATIONS LIMIT EXCEEDED.
ESTIMATION POSSIBLY HAS STALLED OR MACHINE ROUNDOFF IS MAKING FURTHER PROGRESS DIFFICULT
TRY HIGHER SUBITERATIONS LIMIT, TIGHTER CVCRIT, DIFFERENT SETTING FOR EXACTLINE OR ALPHA ON NLPAR
RESTARTING ESTIMATION FROM LAST ESTIMATES OR DIFFERENT INITIAL GUESSES MIGHT ALSO WORK
Dependent Variable PIB
Annual Data From 1960:01 To 2014:01
Usable Observations 55
Degrees of Freedom 50
Centered R^2 0.9984907
R-Bar^2 0.9983699
Uncentered R^2 0.9999862
Mean of Dependent Variable 9.6855946343
Std Error of Dependent Variable 0.9382876296
Standard Error of Estimate 0.0378827609
Sum of Squared Residuals 0.0717551788
Log Likelihood 104.6087
Durbin-Watson Statistic 0.5631
Q(13-2) 76.6459
Significance Level of Q 0.0000000
Variable Coeff Std Error T-Stat Signif
************************************************************************************
1. CONSTANT 7.784659094 0.030028584 259.24163 0.00000000
2. AR{1} -0.754375420 0.035138862 -21.46841 0.00000000
3. MA{1} 0.324101626 0.043120686 7.51615 0.00000000
4. TREND 0.090274713 0.002567841 35.15588 0.00000000
5. TREND2 -0.000598410 0.000045431 -13.17190 0.00000000
I have check many times, I used exaclty the same data. But usable Observations and degrees of fredom are diferents.