Problem with LogL

[Anjuka]

Hi!

I have a disequilibrium model for the loans market (Demand and Supply). Literature suggested to use ML. I have found a program called diseq1.prg in your sample programs which is almost the one I needed. Therefore, I have "copied" your method in solving this type of problem. I have estimated with LS method for the starting values, but the problem appears when I come to LogL. When I type:

@LOGL LOGL1
UD = LOGLOANS-ALPHA(1)-ALPHA(2)*LOGGDP-ALPHA(3)*GDPGAP-ALPHA(4)*INFLATION-ALPHA(5)*LRATELVL
US = LOGLOANS-BETA(1)-BETA(2)*LOGGDP-BETA(3)*LOGEQANDDEP-BETA(4)*DIFFRATES
LOGL1 = -LOG(2*3.14159265358979) -LOG(SIGMA(1)) -LOG(SIGMA(2)) - UD^2/SIGMA(1)^2/2 -US^2/SIGMA(2)^2/2

The following problem appears:
"5 is not a valid index for vector-series-coefficient ALPHA in UD=LOGLOANS-ALPHA(1)-ALPHA(2)*......."

The same goes for:
"4 is not a valid index for vector-series-coefficient ALPHA in US=LOGLOANS-BETA(1)-BETA(2)*......."


I would really appreciate if you can give me a piece of advice on that.
Thanks in advance!

Best regards,
Anna

[EViews Gareth]

You need to have defined a coefficient vector called Alpha (and Beta etc.. for that matter), of the correct size (5 for Alpha by the looks of it).

[Anjuka]

Hi!

Thanks a lot for your respond. Sorry, I did not mention it previously, but actually, I have defined coefficient vectors. Here comes the full version of what I am writing in program (I would really appreciate if you could also point out on some additional problems I might have after running the program (misspecified smth, or missed smth):


coef(5) alpha
coef(4) beta
coef(2) sigma
!pi = @acos(-1)

logl ll1

ll1.append @logl logl1
ll1.append ud = logloans-alpha(1)-alpha(2)*loggdp-alpha(3)*gdpgap-alpha(4)*inflation-alpha(5)*lratelvl
ll1.append us = logloans-beta(1)-beta(2)*loggdp-beta(3)*logeqanddep-beta(4)*diffrates

ll1.append logl1 = -log(2*!pi) -log(sigma(1)) -log(sigma(2)) - ud^2/sigma(1)^2/2 -us^2/sigma(2)^2/2

equation eqa.ls logloans loggdp gdpgap inflation lratelvl
alpha = eqa.@coefs
sigma(1) = eqa.@se
show eqa.output

equation eqb.ls logloans loggdp logeqanddep diffrates
beta = eqb.@coefs
sigma(2) = eqb.@se
show eqb.output

ll1.ml
show ll1.output

[EViews Gareth]

When you're performing the ML estimation, alpha only has 4 rows, since you had set alpha equal to the coefficient vector from equation eqa, and eqa only has 4 regressors. This over-writes the original declaration you made, which made it size 5.

[Anjuka]

Thanks a lot, now it works ;)

I would like also to bother you with one more question regarding the sample program in Eviews 6.0 package - DISEQ1.prg . After running it, I obtain the output for TSLS regressions and ML. If you will recall the original exercise, ML should help in estimating Demand and Supply. How should I proceed from that if I want to get the estimated values for Demand and Supply?

[EViews Gareth]

You should really check the text book that the program/example was taken from.

[Anjuka]

The program (DISEQ1.prg) and example was provided by EViews team :) I've read about it in Users Guide II provided in EViews and the program is in my EViews package.
Therefore, for me it is just a technical question how to get fitted values for estimated Demand and Supply ...

Can you help me on that, please? :)

[EViews Gareth]

That program replicates (as is indicated by the program itself, and the User Guide) an example from a textbook and from a paper. For interpretation of the results etc..., you should refer to the textbook.

[Anjuka]

Thanks for your help, I will check the book one more time!