Some estimates are not produced by the ARIMA function

[John Andrews]

Dear all,

I have estimated an ARMA(2,3) model but some estimates are missing

arima ght , arima(2,0,3)

(setting optimization to BHHH)
Iteration 0: log likelihood = 258.76396
Iteration 1: log likelihood = 263.91729
Iteration 2: log likelihood = 266.25091
Iteration 3: log likelihood = 268.14775
Iteration 4: log likelihood = 268.19946
(switching optimization to BFGS)
Iteration 5: log likelihood = 268.25045
Iteration 6: log likelihood = 268.42298
Iteration 7: log likelihood = 268.5505
Iteration 8: log likelihood = 268.56834 (backed up)
Iteration 9: log likelihood = 268.70447
Iteration 10: log likelihood = 268.73477
Iteration 11: log likelihood = 268.78358
Iteration 12: log likelihood = 268.80113
Iteration 13: log likelihood = 268.80621
Iteration 14: log likelihood = 268.81192
(switching optimization to BHHH)
Iteration 15: log likelihood = 268.81374
Iteration 16: log likelihood = 268.81374 (not concave)
Iteration 17: log likelihood = 268.81377
Iteration 18: log likelihood = 268.81377 (backed up)
Iteration 19: log likelihood = 268.81377 (backed up)
(switching optimization to BFGS)
Iteration 20: log likelihood = 268.81377 (backed up)
Iteration 21: log likelihood = 268.81377
Iteration 22: log likelihood = 268.81379
Iteration 23: log likelihood = 268.8138

ARIMA regression

Sample: 1980q4 thru 1998q2 Number of obs = 71
Wald chi2(4) = 783.95
Log likelihood = 268.8138 Prob > chi2 = 0.0000

------------------------------------------------------------------------------
| OPG
ght | Coefficient std. err. z P>|z| [95% conf. interval]
-------------+----------------------------------------------------------------
ght |
_cons | .0054857 .0016768 3.27 0.001 .0021992 .0087722
-------------+----------------------------------------------------------------
ARMA |
ar |
L1. | 1.678876 .129187 13.00 0.000 1.425674 1.932078
L2. | -.7886234 .1195201 -6.60 0.000 -1.022879 -.5543683
|
ma |
L1. | -1.353673 .2084049 -6.50 0.000 -1.762139 -.9452069
L2. | .2319842 . . . . .
L3. | .4304685 .3080025 1.40 0.162 -.1732053 1.034142
-------------+----------------------------------------------------------------
/sigma | .0053035 .0011699 4.53 0.000 .0030105 .0075964
------------------------------------------------------------------------------
Note: The test of the variance against zero is one sided, and the two-sided
confidence interval is truncated at zero.



Is there any explanation for that?

Best

[John Andrews]

Is there any explanation for this please?