Zero sample correlation does not necessarily mean there is no linear relationship - page 50
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
For a start, it would be good to know how it works.
That's what I mean, to start with, to dig around, to chew on, to understand how he tries to do it.
Allow me to introduce myself: Mathematical Statistics, specialisation:Econometrics.
The matstatistics, and its application to economics econometrics, has a myriad of different kinds of tests, models ..... Each of them is as utterly useless as a spanner without a nut. The usefulness can only be understood within the framework of some objective for which the model is being built.
The discussion here is about WHAT and what is the MODEL of that WHAT? Granger may not be needed.
One thing is clear to me, that only forecasts should be discussed on this forum, while analysis is an aid in building and testing a forecast model.
Then forecast what? price value?, price increment? or something else? Without defining this we will see a set of useless econometric tools and we are bound to be wrong in their interpretation and have different and useless opinions.
Allow me to introduce myself: Mathematical Statistics, specialisation: Econometrics.
...
It's a pleasure.
Now let's get to the point: please describe for us the essence of the Granger test. There's no one to ask. Also, if you could say a few words about transfer entropy.
Allow me to introduce myself: Mathematical Statistics, specialisation: Econometrics.
The matstatistics, and its application to economics econometrics, has a myriad of different kinds of tests, models ..... Each of them is as utterly useless as a spanner without a nut. The usefulness can only be understood within the framework of some objective for which the model is being built.
The discussion here is about WHAT and what is the MODEL of that WHAT? Granger may not be needed.
One thing is clear to me that we should only discuss prediction on this forum, while analysis is an auxiliary tool for building and testing a prediction model.
Then forecast what? price value?, price increment? or something else? Without defining this we will see a set of useless econometric tools and we are bound to misinterpret them and have different and useless opinions.
By the way, while we're philosophizing about what is a method, what is a model, where I(1) and where I(0), I'm riveting another trading system based on this topic:
I mean, maybe we should stop discussing the end of the stick and the beginning, and just take the stick for some end and start making money with it?
I am pleased that you are pleased, but I think you have overestimated me.
I will try to state what remains in my head after the exams.
I don't remember anything about transfer entropy.
About Granger.
There are three somewhat similar notions: correlation, cointegration and Granger test.
Correlation is a constant. If each sample of two SVs for which the correlation is calculated is statistically the same as other samples from the general population of these SVs, then we can say that the two SVs are dependent. More precisely, their behaviour is similar. This holds for normally distributed SV.
If SVs are not normal, then cointegration is applied, when the characteristic of the mutual dependence of two SVs is not a number, but a series with certain properties.
Granger makes it possible to calculate the direction of dependence according to the "chicken or egg first" principle. Another property of dependence.
Here is my understanding on the fingers. But this is a first approximation. Once again. You have to consider it all starting with the description of the original time series, and then the model, and then maybe it will come to the listed concepts.
By the way, while we're philosophizing about what is a method, what is a model, where I(1) and where I(0), I'm riveting another trading system based on this topic:
I mean, can we stop discussing where the stick ends and where it starts, and just take the stick at least by some end and start making money with it?
The problem isn't with ponying up the dough, it's with the understanding that this will be the case tomorrow. Well, with some precision.
Now to cut to the chase: please describe for us the essence of the Granger test
It's simple. An autoregression is estimated on one series, then values taken at different lags from another variable are added to it. Then it is checked whether the second model performs better than the first. If so, there is a causal relationship. Similarly, the autoregression is tested on the second row with inclusion of lag values from the first row. It happens so that the causal relationship is in both directions. It means that the rows are simply correlated.
For transfer entropy read the links from here:
http://stats.stackexchange.com/questions/12573/calculating-the-transfer-entropy-in-r
OK, but what if I understand exactly what my system is working on? I can accurately measure this factor, what more do I need to be happy about?
The ergodicity property allows us to estimate the correlation function for the general population from a sample of the general population. If this property is not fulfilled, the number obtained by the formula can be thrown out.