I would like to share the link - page 8

 
Attached is an interesting article on modelling currency pairs in forex, taking into account trends, variable volatility, cyclicality and news shocks.
 

Attached is a curious article on model evaluation (TC).

The classical approach for TC evaluation is to run it through a tester and get some statistics, and then do a forward test.

The article suggests another approach - through informational criteria different TSs are compared, which may be close due to changes in parameters. By this informational criterion the out-of-sample prediction error is estimated. The author compared 20 million(!) close models and drew some conclusions on this basis.

The interesting thing about the article is precisely the different approach than the usual tradition in TA.

Files:
 

Sometimes a dollar index was used as a synthetic, sometimes rates for each currency were calculated - many different variations.

In the attachment is an article on the same topic using the ruble as an example. I think that the choice of currency does not play a major role. But this article is another look at many of the same topics.

Files:
exchange_ruble.zip  1159 kb
 

Translated an overview of R's time series analysis capabilities. See attachment.

For the lazy, I will list the sections:

Forecasting and Univariate Modeling

Decomposition and Filtering - Decomposition and Filtering

Stationarity, Unit Roots, and Cointegration

Nonlinear Time Series Analysis

Dynamic Regression Models

Multivariate Time Series Models


There is a link to the original. I make no claims on the quality of translation, but it will suffice for informing anyone who wishes to use it.

Please note that there is a huge list of ready-made programs to build TC.

Good luck.

 

Attached is a curious new article on a topic that has been discussed many times on the forum.

It is about the fractal structure of the marketplace. Not many people know that synonymous with the presence of fractals is the presence of long memory in a quotient. The problem of dedernding in such quotients is investigated, and it is argued that one has to deal with the change of scale (temporal) of the quotient, i.e. change of the timeframe - such a mathematical analogue of three windows.

Files:
 
faa1947:

Not many people know that synonymous with the presence of fractals is the presence of long memory in a quotidian.

I will probably anger others if I say that random walk is also fractal and has no memory. Fractality of random walk can be proved mathematically very simple.
 
gpwr:
I would probably anger others if I said that random walk is also fractal and has no memory. Fractality of random walk can be proved mathematically very simple.

It's not about proving something, it's about practical applicability.

With this thread I try to show that there is a great number of theoretical works based on real problems and solving real problems.

To two closely related terms I forgot to add another one which translates the problem of fractals, long memory, thick tails to the practical plane - these are models with fractional integration FARIMA. These are ARIMA models in which the value of I, usually taking positive integers, can be fractional and correspond to the values of Hearst exponent. Hearst has been given much attention on the website, and there is a library in R called fracdiff, a program that solves a number of problems in this area. In addition there are other libraries for dealing with fractals.

I would be very happy to discuss random walk fractality, but subject to the use of some code.

 

R is clearly no slacker.

TIOBE 2012 ranking: R is one of the twenty most popular programming languages

 
Alexey Subbotin:

The nature of the distribution does not change. By the way, the study itself started with the fact that the strange behaviour of Likelihood Ratio is noticeable, one might say, to the naked eye:


Alexey, good day. What is theLikelihood Rat io indicator?