Econometrics: one step ahead forecast - page 19

New comment
 
No, I haven't read it yet, I don't know. But I don't think it's anything like that: I haven't gotten to Chaotica yet. Thanks, I'll download it now.
 
faa1947:

Actually, EViews is one of the most advanced packages in econometrics. Reading additional literature on the state space did not reveal anything that is not in the package.

My offer to cooperate in the state space is still valid.

An example application is in the attachment. But this is not a guide to the program. Find the package, it's with the documentation. Each chapter of the documentation has references to books that outline the relevant algorithms.

The state-space method has a lot of scientific literature -- it has evolved in many directions -- so all its achievements can't be crammed into the econometrics package, no matter how advanced it is... -- rather it uses some elements, nothing more... You can see from the example above that one of the models is being exploited -- but this model in no way reflects the whole picture and capabilities of the method.

I'm not quite clear on your suggestion -- what do you mean by that?

 
avtomat:

There is a lot of scientific literature on the state-space method -- it has evolved in many directions -- so it is unlikely that all the advances could be crammed into one package of econometrics, however advanced it may be... -- rather it uses some elements, nothing more... You can see from the example above that one of the models is being exploited -- but this model in no way reflects the whole picture and capabilities of the method.

I'm not quite clear on your suggestion -- what do you mean?

so all the advances could hardly be crammed into one package of econometrics

Thinking that there was no task to cram everything available in the state space. It's a question of the suitability of the tool to the subject area that the tool serves. In my view the possibilities of models in the state space are immeasurably greater than in regressions, ARMA, ARCH lags almon and the like, which can also be used in the state space. The possibility of modelling some "state" that produces a prediction, the use of "unobservable" events, the extended possibility of modelling residuals ..... - are all more than tempting.

I'm not quite clear on your suggestion -- what do you mean?

Write a model and I'll run it in EViews. Here's an understanding of the model:



If that's enough for you, OK. If not, you have to read the documentation.

 
faa1947:

I'll make another prediction.

The model is unchanged. Opening price at 00:00 on Monday.

Date Value Forecast Value Error R-square Error


Open
at forecast in pips regressions regressions

2011.11.09 00:00 1,383 2011.11.09 1,3798 56 0,9761 55


2011.11.10 00:00 1,3524 2011.11.10 1,3613 60 0,9749 57 -0,0306 -0,0032 correct
2011.11.11 00:00 1,361 2011.11.11 1,3541 59 0,9751 57 0,0086 0,0089 correct
2011.11.14 00:00 1,3778 2011.11.14 1,3676 59 0,9739 57 0,0168 -0,0069 not correct







not known


Forecast Monday, November 14


How is the prediction error calculated in pips? For the first two forecasts, using your data, I get (forecast minus value):

1.3798 - 1.3830 = 32 pips

1.3613 - 1.3524 = 89 pips

 
gpwr:


How do I calculate the forecast error in pips? For the first two forecasts, using your data, I get (forecast minus value):

1.3798 - 1.3830 = 32 pips

1.3613 - 1.3524 = 89 pips

Forecast error is calculated, which I converted into pips (*10000). See last column for formulas
 

faa1947:

so it is unlikely that all the advances could be crammed into one package of econometrics

Thinking that there was no task to cram everything available in the state space. It is a question of the suitability of the tool to the subject area that the tool serves. In my view the possibilities of models in the state space are immeasurably greater than in regressions, ARMA, ARCH lags almon and the like, which can also be used in the state space. The possibility of modelling some "state" that produces a prediction, the use of "unobservable" events, the extended possibility of modelling residuals ..... - are all more than tempting.

I'm not quite clear on your suggestion -- what do you mean?

Write a model and I'll run it in EViews. Here's an understanding of the model:



If that's enough for you, OK. If not, you need to read the documentation.

What do you mean by 'write a model'... The model is given by a system of equations (33.1 - 33.2). It is quite clear and understandable. It takes half an hour to make it in Matkadec (I hope you are not suggesting to make it in EViews). But the other thing is that for linking to existing BP you have to identify model values - you can do it by different methods. The main thing is to understand what you're doing...

By the way, do you understand this model?

 
avtomat:

By the way, do you understand this model?

Unfortunately, not really.

 
faa1947:

Unfortunately, not so much.


econometrics and the right model for trading, Harry's approach is interesting. Spider is not working at the moment, so you can see the discussion thread here
 
Avals:

Harry's approach is interesting in terms of econometrics and the right model for trading. Spider is not working right now, so you can see the discussion thread here
there's no page.
 
faa1947:
no page


Yes, look through Yandex - query "A trader called Harry and his approach to the market" - first link click "copy". Will be from the yandex cache as the spider is now hanging

Or here http://hghltd.yandex.net/yandbtm?fmode=inject&url=http%3A%2F%2Fforex.kbpauk.ru%2Fprintthread.php%2FCat%2F0%2FBoard%2Ftrading%2Fmain%2F39461%2Ftype%2Fthread&text=%D0%93%D0%B0%D1%80%D1%80%D0%B8%20%D1%82%D1%80%D0%B5%D0%B9%D0%B4%D0%B5%D1%80&l10n=ru&mime=html&cht=1&sign=37c0d2a96b68df40eb5368b916539921&keyno=0

1...121314151617181920212223242526...139
New comment