Econometrics: why co-integration is needed - page 17

New comment
 
Farnsworth: In that sense, I've had enough of your last topic and the model with 3 coefficients. How hard it was to convince you that it was bullshit.
You weren't and couldn't convince me because the 3 coefficients model wasn't discussed at all. You didn't even bother to understand what was being discussed. All this time I have tolerated your snobbery and epatage in the hope to see some construction from you, but got just nonsense "There is no ACF stationarity in your series and can't be". I'm sorry for the time wasted on you.
 
faa1947:
Not convinced and couldn't be convinced as the 3 coefficient model was not discussed at all. You didn't even bother to understand what was being discussed. All this time I have tolerated your snobbery and epatage in the hope to see some construction from you, but received just a delirium "There is no ACF stationarity in your series and there cannot be".

Oh, boy! Constructive? You ask me where your money is and I'll tell you where it is. Your 3 coefficients were discussed earlier, I just remembered them. And that is what is being discussed here, the main thing is that you understand it. And you don't have ACF stationarity - that's a fact. If you weren't so stubborn, you would have checked and verified it.

I'm sorry for the time wasted on you.

Already wrote in the last thread, the doctor is available 24 hours a day, 5 days a week: from 01:00 (msec) Monday to 01:00 (msec) Saturday.

 

to faa

I won't give statistics of stationarity estimation, there's no sense, but in addition here is ACF of the transformation process (increments I work with the first 500 samples (it is EURUSD, M15). The black and grey lines are ACFs shifted by two trading weeks.

And this, the time shift of the ACF is already a few trading months away.

And there are real stationarity problems, all statistics are at the "limit" and the transformation itself is very complex, sharpened for stationarity reduction based on wavelets. And you again "regress" on a few coefficients, look in EW, interpret the "physics" data quite strangely and naively think that these new coefficients of yours will give a good prediction.

OK, I'll waste serious time on you too, I'll be like paukas, - joking :o)

 
anonymous: http://www.burns.com/wcbspurcorl.htm

False correlation (not in the graph, but in the caption below it):


 
Mathemat:

False correlation (not in the graph, but in the caption below it):

Nah... As they say, advertising is not an offer, but only a call to make offers :)

False will be when they state interdependence.

By the way, there is no such thing as a false correlation - just as there is no such thing as a false hypothesis or a false decision.

 
Farnsworth:


Sergey, why don't you want to accept the author's position and confirm it with arguments, refute it, or just discuss it?

Put out the required data (quote and balance with constant time step) of your own EA, get SanSanych's assessment and trash it :)

 
tara: False would be when interdependence is stated.

No one said there had to be interdependence. One-way dependence is enough.

The false correlation is: "there is a direct correlation between megacities and Poo's tenure in power". Or now this: "the more Pu is in power, the higher the megacits on average".

What's wrong with that?

 

There are no false correlations - just as there are no wrong hypotheses or wrong decisions.

 

It's like centrifugal force. Which also doesn't happen :)

 

tara: ложных корреляций не бывает,- так-же, как ошибочных гипотез и неправильных решений.

It happens, it happens. Especially by British scientists.
1...101112131415161718192021222324...28
New comment