Graph Condtional Moments

[TomDoan]

The second and third terms. With the "temp" variables being zero, the third washes, so it was the 2nd terms that didn't match. Your fix takes care of the second term, but you probably want to fix the third as well in case you ever use a non-zero set of "temp" series.

[rbelhach95]

Thanks so much Tom.

Does this looks reasonable now s1 = beta10*lambda10+beta11*lambda11*(e{1}-gama11)*(e{1}-gama21)+beta12*lambda12*temp2{1}*temp4{1}+(alpha11*alpha31)*f1{1}^2+(alpha12*alpha32)*f2{1}^2,

Very much appreciated
Rachid

[rbelhach95]

Sorry I meant sqrt in both temp2 and temp4.
Thanks again
Rachid

[TomDoan]

That looks correct. Note, BTW, that you're asking us to check your algebra, not your RATS program, which is really outside the scope of what we should be doing.

[rbelhach95]

Sorry about that.
Thanks

[rbelhach95]

Dear Tom,

Hope all is well. Regarding the two attached codes in my previous email (code for parameter estimation and graphs of conditional moments) is there a way to combine both codes in one so that I don't have to each time do copy and paste of the parameters I get from the first code into the graph code. Thanks
Rachid

[TomDoan]

Of course. What you did is harder than combining them.

[rbelhach95]

Thanks for your prompt reply, well how is it done then? Are there e.g internal name variables for these parameters, and if so how are they defined, I am willing to find out myself how it is done, if you could direct me to the RATS user guide for an example. Thanks for your support
Rachid

[TomDoan]

Just take the part of your graphing program that does the graphs given the parameters, and add it to the end of the estimation program. One copy and paste, one time. Done.

[rbelhach95]

Thanks Tom, i get the point. Very much appreciated
Rachid

[rbelhach95]

Dear Tom,

I don't really know what I am doing wrong here, but I did exactly what you suggested on your last email, when I run the codes separately I get nice graphs, when I join them I get rubbish or an error by "set e". I attached the joined codes. Please help
Thanks
Rachid

[moderator]

In the first part, you have the following definitions as FRML's so you can estimate:

frml e = x-a10-a11*x{1}-a12*temp1{1}
frml mux = a10+a11*x{1}+a12*e{1}
frml muT = a20+a21*f1{1}+a22*f2{1}
frml d1 = beta10^2+(beta11^2)*(e{1}-gama11)^2+(beta12^2)*temp2{1}+(alpha11^2)*f1{1}^2+(alpha12^2)*f2{1}^2
frml d2 = beta20^2+(beta21^2)*(e{1}-gama12)^2+(beta22^2)*temp3{1}+(alpha21^2)*f1{1}^2+(alpha22^2)*f2{1}^2
frml k1 = lambda10^2+(lambda11^2)*(e{1}-gama21)^2+(lambda12^2)*temp4{1}+(alpha31^2)*f1{1}^2+(alpha32)^2*f2{1}^2
frml k2 = lambda20^2+(lambda21^2)*(e{1}-gama22)^2+(lambda22^2)*temp5{1}+(alpha41^2)*f1{1}^2+(alpha42^2)*f2{1}^2
frml sigmaTx1 = (beta10*lambda10)+(beta11*lambda11)*(e{1}-gama11)*(e{1}-gama21)+(beta12*lambda12)*temp6{1}+(alpha11*alpha31)*f1{1}^2+(alpha12*alpha32)*f2{1}^2
frml sigmaTx2 = (beta20*lambda20)+(beta21*lambda21)*(e{1}-gama12)*(e{1}-gama22)+(beta22*lambda22)*temp7{1}+(alpha21*alpha41)*f1{1}^2+(alpha22*alpha42)*f2{1}^2

You're then trying to re-define them as SERIES in the second half. You don't need to do that. Just get rid of the SET's for all the series covered by that set of FRML's.

[rbelhach95]

Dear Tom,

Thanks a lot for your continuous support. I have a cleaned up the code, please see the attached edited "joined code_1", however, when I run this code I still get rubbish graphs as opposed to good graphs when run separately. I used the "print instruction" to scrutinize my output, and I see that in the joined code many data points are either NA or incorrectly calculated, but for the two separate codes I get good values. Please advise.
Thank you
Rachid

[rbelhach95]

Please see the 2 attached graphs mentioned in my last email
Thanks
Rachid

[TomDoan]

Dear Tom,

Thanks a lot for your continuous support. I have a cleaned up the code, please see the attached edited "joined code_1", however, when I run this code I still get rubbish graphs as opposed to good graphs when run separately. I used the "print instruction" to scrutinize my output, and I see that in the joined code many data points are either NA or incorrectly calculated, but for the two separate codes I get good values. Please advise.
Thank you
Rachid

Isn't this a different model? Perhaps you are getting bad graphs because you're getting bad estimates. These don't look usable to me. Among other things, it's a very bad idea to use lambda21=0 as a guess value for (lambda21^2)*(e{1}-gama22)^2 because gama22 isn't identified at the guess values and the derivative with respect to lambda21 is zero.

Code: Select all

MAXIMIZE - Estimation by BFGS
NO CONVERGENCE IN 114 ITERATIONS
LAST CRITERION WAS  0.0000000
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
Monthly Data From 1978:12 To 2013:09
Usable Observations                       413
Skipped/Missing (from 418)                  5
Function Value                        NA

    Variable                        Coeff      Std Error      T-Stat      Signif
************************************************************************************
1.  A10                             1.0473178    0.2253669      4.64717  0.00000337
2.  A11                            -0.4866274    0.0785399     -6.19593  0.00000000
3.  A12                             0.5360456    0.0797345      6.72288  0.00000000
4.  A20                             1.5161977    1.2180825      1.24474  0.21322681
5.  A21                            -4.4407812    1.2286561     -3.61434  0.00030111
6.  A22                             0.2974692    0.0092911     32.01654  0.00000000
7.  BETA10                         -0.0415601    0.5524054     -0.07523  0.94002793
8.  BETA11                          0.2634283    0.0657946      4.00380  0.00006233
9.  BETA12                          0.9224623    0.0210704     43.78003  0.00000000
10. BETA20                         -0.2974384    0.3603854     -0.82533  0.40918184
11. BETA21                          0.4491172    0.0576414      7.79158  0.00000000
12. BETA22                         -0.3750155    0.1160372     -3.23186  0.00122989
13. LAMBDA10                        0.2063586    0.0745827      2.76684  0.00566021
14. LAMBDA11                        2.8929021    0.6782412      4.26530  0.00001996
15. LAMBDA12                       -2.3274169    0.4034412     -5.76891  0.00000001
16. LAMBDA20                       -0.3384745    0.8577936     -0.39459  0.69314746
17. LAMBDA21                        0.0009211    0.0024188      0.38079  0.70335944
18. LAMBDA22                        0.4618360    0.1967321      2.34754  0.01889797
19. ALPHA11                         0.4083553    0.4547312      0.89801  0.36917773
20. ALPHA12                        -0.0052590    0.0038077     -1.38116  0.16723030
21. ALPHA21                        -1.6911879    0.4783948     -3.53513  0.00040757
22. ALPHA22                        -0.0246184    0.0040149     -6.13169  0.00000000
23. ALPHA31                         0.8256964    0.4329012      1.90736  0.05647458
24. ALPHA32                         0.0624921    0.0032536     19.20685  0.00000000
25. ALPHA41                        -1.9729743    0.9038829     -2.18278  0.02905229
26. ALPHA42                         0.0021609    0.0098981      0.21832  0.82718236
27. GAMA11                          1.1373636    1.1658485      0.97557  0.32927901
28. GAMA12                          7.2798033    0.9785560      7.43933  0.00000000
29. GAMA21                       -297.9302882   70.0962469     -4.25030  0.00002135
30. GAMA22                       -329.2616947  127.3408296     -2.58567  0.00971892