model newton method

For technical questions regarding estimation of single equations, systems, VARs, Factor analysis and State Space Models in EViews. General econometric questions and advice should go in the Econometric Discussions forum.

Moderators: EViews Gareth, EViews Moderator

deroekonom
Posts: 8
Joined: Wed Nov 26, 2008 6:59 pm

model newton method

Postby deroekonom » Fri Dec 05, 2008 3:30 am

hello,

I'am not very familiar with evies. However, I have a complex system of thirty equations with the 30 unknown variables delta_i which is are on the rhs of the equation.

For all the other variables in the equation system I have empirical values, and I know that I can get "real values" for the 30 deltas by using the Newton method. But I have almost no clue how to do that in eviews.
I know that I can bring all the data by using "empty group" into eviews. And that there I can start a new object or model, and that I have to write down the equation there. However, that is the first problem I have. I don't know how to write down the equation (is there some formel-editor? For example I need to write down a sigma sign, etc.).

And there is another problem with the empirical data: partly it is panel data, partly not. Is that a problem?

I'am very thankful to every answer/help I can get.
Thank you.
doe

deroekonom
Posts: 8
Joined: Wed Nov 26, 2008 6:59 pm

Re: model newton method

Postby deroekonom » Thu Feb 19, 2009 4:17 am

Is there a special reason why I get no answer?

A more specific question:
For the deltas which I described above, how can I assign them "starting values" (i.e. I have some idea of their range).

Thank you so much for any advice.

EViews Gareth
Fe ddaethom, fe welon, fe amcangyfrifon
Posts: 12261
Joined: Tue Sep 16, 2008 5:38 pm

Re: model newton method

Postby EViews Gareth » Thu Feb 19, 2009 9:25 am

I'm not sure I follow your question at all. Could you reword it perhaps?

I understand that you have managed to get some data into EViews, and that you wish to solve some sort of estimation with that data. After that, I'm lost...
Follow us on Twitter @IHSEViews

deroekonom
Posts: 8
Joined: Wed Nov 26, 2008 6:59 pm

Re: model newton method

Postby deroekonom » Tue Feb 24, 2009 12:37 am

hi,

thank you for your answer.

Here is my problem again:

I have the following equation x = a + b y + Σ(i=1,..31, i≠j) delta/z_ij e^(s_ij/delta - 1)^2 + t

a and b are scalars, s_ij is matrix, z_ij is a matrix, b and t are vectors, delta is the unkown (for all the other variables I have empirical values).

I know that I can get "real values" for delta by using the so-called "Newton method".
If I build a model in e-views, you can choose between the Newton method and the Gauss-Seidel method to solve the model.

However, I'am not sure how to implement the equations in the model (simply technically, e.g. I don't know how to assign the indices of the vectors and matrices).
Second, I'am not 100% sure wheather e-views can deal with my probleam at all.

Thank you for any advice you have.

deroekonom
Posts: 8
Joined: Wed Nov 26, 2008 6:59 pm

Re: model newton method

Postby deroekonom » Tue Feb 24, 2009 1:26 am

to clarify the issue of the index:

If I write for example in the text window of models the following:

a(1) to indicate the first element of vector a, I get the folowing error message:
"(1) is not a valid index for in "the equation written in the text window"

if I simply write: a (without an index), then I get the following error message
"0 is not a valid index for vector-series-coefficient A in "the equation written in the text window"

So, I'am confused.
How can I assign indices to vectors in e-views?

EViews Gareth
Fe ddaethom, fe welon, fe amcangyfrifon
Posts: 12261
Joined: Tue Sep 16, 2008 5:38 pm

Re: model newton method

Postby EViews Gareth » Tue Feb 24, 2009 9:10 am

EViews can't really perform that sort of estimation. You'd be better off with some sort of matrix language with an optimiser - such as Matlab or Gauss.
Follow us on Twitter @IHSEViews

deroekonom
Posts: 8
Joined: Wed Nov 26, 2008 6:59 pm

Re: model newton method

Postby deroekonom » Wed Feb 25, 2009 7:52 am

Thank you for your advive:

A last question: Are you 100% sure that e-views can't perform my problem?
I just ask because there is a "Newton solver" in the "model part" of e-views.
However, I'am not sure if this is THE newton solver I'am looking for or it this one is something different.

Thank you so much again.

EViews Gareth
Fe ddaethom, fe welon, fe amcangyfrifon
Posts: 12261
Joined: Tue Sep 16, 2008 5:38 pm

Re: model newton method

Postby EViews Gareth » Wed Feb 25, 2009 9:13 am

Models (and all Nonlinear estimation, I believe) in EViews do use the Newton Solver, and it is probably the Newton Solver you're talking about. However you still cannot perform the sort of estimation you want, because there is no way to structure the problem in EViews.
Follow us on Twitter @IHSEViews


Return to “Estimation”

Who is online

Users browsing this forum: No registered users and 5 guests