BOXJENK output

[timduy]

Working on an example of what happens when you model a nonstationary series as trend stationary. Running this code:

Code: Select all

boxjenk(constant,ar=1,ma=1,define=gdpmodel,regressors,maxl) lgdp startdate enddate
#trend

This is the output:

Code: Select all

Box-Jenkins - Estimation by ML Gauss-Newton
Convergence in    16 Iterations. Final criterion was  0.0000067 <=  0.0000100

Dependent Variable LGDP
Quarterly Data From 1995:01 To 2011:04
Usable Observations                        68
Degrees of Freedom                         64
Centered R^2                        0.9976231
R-Bar^2                             0.9975117
Uncentered R^2                      0.9999996
Mean of Dependent Variable       9.5225398422
Std Error of Dependent Variable  0.1281958299
Standard Error of Estimate       0.0063948371
Sum of Squared Residuals         0.0026172123
Log Likelihood                       247.3154
Durbin-Watson Statistic                1.7413
Q(17-2)                               23.8108
Significance Level of Q             0.0683775

    Variable                        Coeff      Std Error      T-Stat      Signif
************************************************************************************
1.  CONSTANT                     8.0812375566 0.2014403560     40.11727  0.00000000
2.  AR{1}                        0.9761351790 0.0294054725     33.19570  0.00000000
3.  MA{1}                        0.2822992057 0.1237816244      2.28062  0.02590984
4.                               0.0062448958 0.0009017912      6.92499  0.00000000

Why is variable 4 not identified as "trend"?

Thank you.

[TomDoan]

I don't know---I've never seen something like that. Could you give us the whole program?

[timduy]

Attached.