Machine learning in trading: theory, models, practice and algo-trading - page 2731
You are missing trading opportunities:
- Free trading apps
- Over 8,000 signals for copying
- Economic news for exploring financial markets
Registration
Log in
You agree to website policy and terms of use
If you do not have an account, please register
Statistical models are GARCH models.
Take the rugarch package and enjoy life? Everything is chewed up.
In brief.
Statistical models are based on the assumption that financial markets are non-stationary. That is why all ordinary statistics, including the above mentioned tests, goes in the basket.
Therefore:
1. Financial series are tedrending, usually increments are taken (very good for us).
There are publications that IGARCH models are the most suitable for financial markets
The statistical models are GARCH models.
Take the rugarch package and enjoy life? Everything is chewed up.
To summarise.
Statistical models are based on the assumption that financial markets are non-stationary. Therefore, all the usual statistics, including the above mentioned tests, go in the basket.
All autoregressive models are some transformations of white noise. And if the inverse transformation does not yield white noise in the end, the model also goes to the basket. And white noise is first of all a stationary process.
This is a very important point in statistical models - any modelled nonstationarity is based on stationarity, which actually gives the very possibility to study it.
Examples with usage?
Google to the rescue. There's a huge literature.
All autoregressive models are some transformations of white noise. And if the inverse transformation does not yield white noise in the end, the model also goes to the bin. And white noise is first of all a stationary process.
This is a very important point in statistical models - any modelled nonstationarity is based on stationarity, which actually gives the very possibility to study it.
Read the Garch and don't make things up
Google to the rescue. There's a huge literature.
Question for you, not google.
It's not you, it's YOU.
Without me, talk to your own kind.
Nuance: the operation of "detrending" kills all the most tasty and interesting :-)
More precisely, there is simply no reasonable way to remove trends or everything but trends. There are methods to minimise or take into account seasonals (daily/weekly fluctuations). Perhaps the big personalities have methods for months and quarters.
Just the very concept of "trend" is so vague that it exists only in the head when viewing the accomplished history. But big/small trends are traded.
----
Right now, in the objectively observed reality, there is an obvious trend of USD growth against all. More precisely, de facto and for a long time, it is going on. And it is very likely that today-tomorrow it is over, because it is obvious.
That is, the fact of growth takes place. Everyone determines the moment of the beginning of the growth differently and only after a considerable time (someone probably just drained his account against the quid, that's why he noticed it).
Is there a method that will determine in time "Oh! big trend...it is necessary to remove its influence from small ones"? And there is no such method. Not without looking ahead.
----
The very trading task on a single instrument is to determine the trend in time. And there is a dead end here, "trend" :-) there is no generally accepted technical definition that can be reduced to the level of formulas and algorithms (they are basically just another way of writing formulas).
SUMMARY: we should pay attention to the operation "detrending". On the one hand, there is no way to do without it, on the other hand, too much necessary information dies there.
Not YOU, but YOU.
Without me, socialise with your own kind.
How many of you are there?
You have nothing to say about the question you're asking:
1) think of something to pick on
2) take offence
3) piss off