[Limdep Nlogit List] Problem with NLSUR

Rao, James James.Rao at agr.uni-goettingen.de
Thu May 28 20:02:27 EST 2009


Hi,

Below is an extract from an iterative procedure and including some errors I
encountered
Even though the the iterations converged, I ended up with a flat log
likelihood and I am wondering what could have caused this and if there is any
possible solution to this problem!

*****************************************************************************
**********
Itr 35 F=  .9732D+02 gtHg=  .2105D-05 chg.F=  .2857D-08 max|db|=  .1669D-01
Try =  0 F=  .9732D+02 Step=  .0000D+00 Slope= -.2394D-04
Try =  1 F=  .9732D+02 Step=  .9903D-05 Slope=  .1288D-02
Try =  2 F=  .9732D+02 Step=  .1807D-06 Slope=  .1802D-09
1st derivs.     .40415D-07   .13522D-06   .16210D-06   .19216D-06
-.68662D-07
  -.18281D-06  -.38962D-06  -.11943D-05   .42140D-05  -.66719D-07
-.17864D-06
  -.15762D-06  -.23054D-06  -.32605D-06  -.80907D-05   .68696D-07
-.14971D-06
   .21338D-05   .31859D-06  -.20567D-06  -.58567D-07   .00000D+00
.00000D+00
  -.28370D-06   .00000D+00   .26538D-06   .36984D-06
Parameters:     .97155D+01  -.11219D+00  -.12054D+00  -.92209D-01
-.28115D+00
   .94535D-01   .43009D-01   .21913D+00   .60080D-01  -.12993D+00
.40283D-01
  -.37693D-01   .16186D+00   .41855D+00   .28355D-02   .24691D+00
-.29970D+00
   .71434D-02  -.92272D+00  -.18516D+00  -.24134D+00   .20226D+01
-.15615D+01
  -.50900D-09  -.89078D+00   .27635D+01   .68991D-01
Itr 36 F=  .9732D+02 gtHg=  .9592D-07 chg.F=  .2331D-11 max|db|=  .5039D-03
                        * Converged
Normal exit from iterations. Exit status=0.
Function=  .16166515574D+04, at entry,  .97323833448D+02 at exit
Iteration=  1, log|S|=     -.6235204, ||db||=     4.7172716
Iteration=  2, log|S|=     -.6277980, ||db||=      .0141203
Iteration=  3, log|S|=     -.6279388, ||db||=      .0003040
Iteration=  4, log|S|=     -.6279496, ||db||=      .0000062
Iteration=  5, log|S|=     -.6279510, ||db||=      .0000001
Iteration=  6, log|S|=     -.6279512, ||db||=      .0000000
  Error   806: (The log likelihood is flat at the current estimates.)
Iteration=  7, log|S|=     -.6279512, ||db||=      .0000000
  Error   805: Initial iterations cannot improve function.Status=3
Iteration=  8, log|S|=     -.6279512, ||db||=      .0000000
  Error   143: Models - estimated variance matrix of estimates is singular
  Error   447: Current estimated covariance matrix for slopes is singular.
*****************************************************************************
*****
=============================================================================
==
Rao E.O. James (Msc)
Research Associate/PhD Student

Department of Agricultural Economics and Rural Development
Georg-August University of Göttingen
Platz der Gottinger Sieben 5
37073 Göttingen, Germany.
 
Tell:+49-551-394443 (Office)
E-mail: jimrao2 at yahoo.co.uk;  James.Rao at agr.uni-goettingen.de

Courage is not the absence of fear but the triumph over it ... so fear not to
fear ... but strive to overcome fear!




-----Original Message-----
From: limdep-bounces at limdep.itls.usyd.edu.au on behalf of William Greene
Sent: Sat 5/16/2009 7:28 AM
To: Limdep and Nlogit Mailing List
Subject: Re: [Limdep Nlogit List] Problem with NLSUR
 
Mr. Rao.
(1) The warning about the log likelihood is harmless if the derivatives
are small and you have made more than a few iterations. In your case,
it sounds like it can be ignored.
(2) NLSUR iterates over the covariance matrix outside the iterations 
for the parameters. Each time it "jumps back" it is updating the 
disturbance covariance matrix and reentering the iterations.
B. Greene

----- Original Message -----
From: "James Rao" <James.Rao at agr.uni-goettingen.de>
To: "Limdep and Nlogit Mailing List" <limdep at limdep.itls.usyd.edu.au>
Sent: Friday, May 15, 2009 12:32:19 PM GMT -05:00 Columbia
Subject: Re: [Limdep Nlogit List] Problem with NLSUR


Hi,

Back to my usual problem with NLSUR ...... my iterations converged well, but
there are two or three lines of error messages saying that "The log
likelihood is flat at current estimates". What could be the implications of
this error? Does it mean that the iterations did not actually converge? And
what could be a potential solution to this problem?

The second concern is that the iterations proceed smoothly but then jump back
to first iteration after some iterations .... say 15 iterations. What could
be the cause of these jumps?

Sincerely
=============================================================================
==
Rao E.O. James (Msc)
Research Associate/PhD Student

Department of Agricultural Economics and Rural Development
Georg-August University of Göttingen
Platz der Gottinger Sieben 5
37073 Göttingen, Germany.
 
Tell:+49-551-394443 (Office)
E-mail: jimrao2 at yahoo.co.uk;  James.Rao at agr.uni-goettingen.de

Courage is not the absence of fear but the triumph over it ... so fear not to
fear ... but strive to overcome fear!




-----Original Message-----
From: limdep-bounces at limdep.itls.usyd.edu.au on behalf of Rao, James
Sent: Tue 4/28/2009 5:27 PM
To: Limdep and Nlogit Mailing List
Subject: Re: [Limdep Nlogit List] Problem with NLSUR
 
Hi, 

In estimating the a non linear systems of equations am using the NLSUR
command that corresponds to "nonlinear IV systems of equations (NL3SLS)". In
the table of results shown below, I wonder whether or not its right that the
"Number of parameters is shown to be = 0"

Is it also right that my information criteria are all the same?

+---------------------------------------------+
| Nonlinear minimization over  3 equations.   |
| Maximum Likelihood Estimates                |
| Model estimated: Apr 24, 2009 at 03:17:54PM.|
| Dependent variable             MultEqns     |
| Weighting variable                 None     |
| Number of observations              402     |
| Iterations completed                 17     |
| Log likelihood function        101.8690     |
| Number of parameters                  0     |
| Info. Criterion: AIC =          -.50681     |
|   Finite Sample: AIC =          -.50681     |
| Info. Criterion: BIC =          -.50681     |
| Info. Criterion:HQIC =          -.50681     |
| Disturbances are   correlated               |
| Covariance matrix used is (1/N)E'E          |
| Number of iterations over S is    11        |
| Used multiple    equation nonlinear 3S-IV.  |
| McElroy R-squared for the system =  .46612  |
+---------------------------------------------+


=============================================================================
==
Rao E.O. James (Msc)
Research Associate/PhD Student

Department of Agricultural Economics and Rural Development
Georg-August University of Göttingen
Platz der Gottinger Sieben 5
37073 Göttingen, Germany.
 
Tell:+49-551-394443 (Office)
E-mail: jimrao2 at yahoo.co.uk;  James.Rao at agr.uni-goettingen.de

Courage is not the absence of fear but the triumph over it ... so fear not to
fear ... but strive to overcome fear!


_______________________________________________
Limdep site list
Limdep at limdep.itls.usyd.edu.au
http://limdep.itls.usyd.edu.au

_______________________________________________
Limdep site list
Limdep at limdep.itls.usyd.edu.au
http://limdep.itls.usyd.edu.au
_______________________________________________
Limdep site list
Limdep at limdep.itls.usyd.edu.au
http://limdep.itls.usyd.edu.au




More information about the Limdep mailing list