diferent trace in function "Estimate"

[imendez]

I think the function 'Estimate' is different in the CVS from version 1.1.10a
(2004/03/18 update). In this version the trace is:

Iteración( 0) norma : 0.044315 Tiempo : 0.360000 segundos

Evaluación ARIMA : (stdErr = 0.044000) * (|ACov|^{(1/2N) = 1.005312) * (Sqrt}

((w'*InvACov*w/A'A) = 1.001825); stdev(difNoise) = 0.062228)

Marquardt regularization (1) with lambda=0.010000 norma 0.033822
Marquardt regularization (2) with lambda=0.002500 norma 0.033819

Iteración( 1) norma : 0.033818 - 0.044315 = -0.010496 (-23.686%)
maxDifx=0.860794 Max|Jta|=46.598638 Tiempo : 139.079000 segundos

Evaluación ARIMA : (stdErr = 0.033551) * (|ACov|^{(1/2N) = 1.004885) * (Sqrt}

((w'*InvACov*w/A'A) = 1.003074); stdev(difNoise) = 0.049840)

Marquardt regularization (1) with lambda=0.002500 norma 0.033795
Marquardt regularization (2) with lambda=0.000625 norma 0.033795

Conjugate direction [1]=001.0% norm=0.033818
Conjugate direction [2]=001.8% norm=0.033817
Conjugate direction [3]=002.6% norm=0.033817

Iteración( 2) norma : 0.033817 - 0.033818 = -0.000001 (-0.004%)
maxDifx=0.105528 Max|Jta|=1.786230 Tiempo : 143.440000 segundos
Tiempo Método de minimización cuadrática de Marquardt 4 minutos 43.254
segundos

Evaluación ARIMA : (stdErr = 0.033549) * (|ACov|^{(1/2N) = 1.004897) * (Sqrt}

((w'*InvACov*w/A'A) = 1.003065); stdev(difNoise) = 0.049830)
2004/03/23 22:23:43 :

However, in the CVS version doesn´t appear, for instance, the "Conjugate
direction " trace.

[danirus]

Estimation Engine Bug Report

[manuelb]

Moved to the new component "ESTIMATE".

[Jorge]

Reassigned to Victor

[Víctor de Buen Remiro]

Dear Ivan

Sorry, but I've estimated a simulated ARIMA model and the Conjugate Gradient
traces are already well showed.

Please, ensure you are using just the last tol-project cvs version. If error persists
you must send here the correspondent TOL code in order to reproduce the bug;
else you must notify it to end this bug's thread.

Thanks and sorry for my poor english.
Best regards

[Víctor de Buen Remiro]

I've received more information from Ivan about this problem and now I'm sure
that the used version is not the tol-project CVS version but another intermediate
old version.

The traces are

MaxLH = 0.044684 = (stdErr = 0.044446) * (|aCov|^{(1/2N) = 1.005312)}

• (Sqrt((R'SR/A'A) = 1.000034); stdev(difNoise) = 1.669648)

Iteración( 0) Norma : 0.044684 Tiempo : 8.110000 segundos

MaxLH = 0.039099 = (stdErr = 0.033304) * (|aCov|^{(1/2N) =}

1.005323) * (Sqrt((R'SR/A'A) = 1.167787); stdev(difNoise) = 0.060475)
Iteración( 1) Norma : 0.039099 - 0.044684 = -0.005585 (-12.499%) : Tiempo :
410.333000 segundos

MaxLH = 0.040000 = (stdErr = 0.033276) * (|aCov|^{(1/2N) =}

1.005653) * (Sqrt((R'SR/A'A) = 1.195292); stdev(difNoise) = 0.062195)

MaxLH = 0.039972 = (stdErr = 0.033276) * (|aCov|^{(1/2N) =}

1.005650) * (Sqrt((R'SR/A'A) = 1.194454); stdev(difNoise) = 0.062131)

Subiteración( 2, 1) Lambda : 0.021124 Norma : 0.039972 - 0.039099 =

0.000873 ( 2.232%) : Tiempo : 364.817000 segundos

MaxLH = 0.039886 = (stdErr = 0.033276) * (|aCov|^{(1/2N) =}

1.005644) * (Sqrt((R'SR/A'A) = 1.191901); stdev(difNoise) = 0.061968)

Subiteración( 2, 2) Lambda : 0.042248 Norma : 0.039886 - 0.039099 =

0.000787 ( 2.012%) : Tiempo : 365.114000 segundos

MaxLH = 0.039666 = (stdErr = 0.033276) * (|aCov|^{(1/2N) =}

1.005621) * (Sqrt((R'SR/A'A) = 1.185351); stdev(difNoise) = 0.061549)

Subiteración( 2, 3) Lambda : 0.084497 Norma : 0.039666 - 0.039099 =

0.000567 ( 1.449%) : Tiempo : 363.021000 segundos

MaxLH = 0.039368 = (stdErr = 0.033276) * (|aCov|^{(1/2N) =}

1.005555) * (Sqrt((R'SR/A'A) = 1.176543); stdev(difNoise) = 0.060986)

Subiteración( 2, 4) Lambda : 0.168993 Norma : 0.039368 - 0.039099 =

0.000270 ( 0.689%) : Tiempo : 364.285000 segundos

MaxLH = 0.039180 = (stdErr = 0.033277) * (|aCov|^{(1/2N) =}

1.005450) * (Sqrt((R'SR/A'A) = 1.170985); stdev(difNoise) = 0.060635)

Subiteración( 2, 5) Lambda : 0.337986 Norma : 0.039180 - 0.039099 =

0.000081 ( 0.207%) : Tiempo : 364.474000 segundos
Iteración( 2) Norma : 0.039099 - 0.039099 = 0.000000 ( 0.000%) : Tiempo :
2229.154000 segundos
Tiempo Método de minimización cuadrática de Marquardt 44 minutos 7.597
segundos

and this is now an obsolete method of traces.