Regression equation - page 5

New comment
 
very clearly, with the iconic method being only the simplest thing you can think of.
 

Again, it is a question of the adequacy of the chosen model.

FreeLance I think that if the number of degrees of freedom is not difficult and the type of dependence is clear a priori - what is the question?

Do you understand the dependence a priori? I don't.

to alsu


The question here is not about the degree, but which polynomial to choose. For example according to Draper Smith, for time series increasing degree of Chebyshev polynomial above 3 may lead to significant worsening of results.

And there is still the question of multivariate regression - involving all available "measurement" data.


One can, IMHO, speak of MNC in the case of an advective model. Is there one? !!!!

P.S. Why did I pick on the multivariate model? Yes because according to some studies the efficiency (predictive) increases considerably.

 

j21, put your research out there: there's already an interesting discussion going on. And there are a lot of interesting people here.

P.S. Your thread threatens to become one of the most entertaining and informative on this forum. There are few of them, really few.

 
Yeah, j21, go ahead. I've been dabbling in multivariate regression as well, we'll discuss it. Tomorrow:)
 

Indirect reference to the article. Scientific publication: Zhdanov A.I., Muravyev D.G. "About one regression method of currency quotes forecasting" (Samara).


I personally have difficulties with formalization.

 

Holy shit.

This is exactly what I was talking about: the beginning of the article - P(z) is unknown, the end of the article - we use t-criterion, i.e. we assume normal distribution. It turns out that the author is simply pasting data into formulas without understanding what implicit assumptions are made in the process.

 
here are my five cents on the subject. You are advised to read http://reslib.com/book/26864
 

Thank you! By the way, reslib has quite a lot of scientific literature on applied linear regression and regression equations in conjunction with time series. Unfortunately there is a page limit on reslib. Unfortunately the book is in english, it is a bit difficult to read.

Can you give the main gist of the question of multivariate regression from a time series perspective?

 

Regarding the article - I have seen somewhere an implementation (or similar) of the algorithm (by these authors). As soon as I find it - I'll post it.

P.S. There is no full text of the article. ((

 
alsu:

least distances method or quantile regression



Can you go into more detail about this, or a link to where to find it.
123456789101112...16
New comment
Reason: