Econometrics: one step ahead forecast - page 60
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
Where did you get the idea that I was considering the stationarity of the market? I have been on the forum for many years and have always said that the market is non-stationary.
The big fans of stationarity are the portfolioists with their efficient market. There are many such people, I think the vast majority, but I don't belong to them.
Non-stationarity is not a property of the world, it is a property of observing its phenomena. More precisely, when observations of completely different phenomena are mixed into one pile. For example, we observed temperature change in Africa, and then started measuring it in Antarctica, and then generally humidity in Brazil and presented it all in a single number series. Just because we have prices does not mean that there is one process behind their formation - there are several completely different and sometimes opposite manifestations.
Bifurcations of course take place, but there are also stable sections. Without them there would be no bifurcation since it is a manifestation of contradictions accumulated in a stable section. Purely from life, you can compare it with revolutions: as long as the political system is stable it cannot be changed even by strong influences, but people's dissatisfaction with life is accumulating and when it has reached critical values, even a minor event can lead to a situation that will develop in this or that scenario. A student will be locked up by the police and mass protests and riots will ensue. Or maybe they won't, and everything will remain the same for a few days/months/years. It's the same in the market - a bifurcation is preceded by a steady stretch with an accumulation of contradictions. For example, the accumulation of unrecorded profits by the traders-intraders. And of course the bifurcations and the processes that precede them can be of absolutely different scales.
What is the point of all this? We have to filter the bullshit. I mean the market :)
I like your reasoning very much. Everything we see around us is not the consequence of exactly one cause. A person who applies matstatistics should always keep this in mind. The basic concept of correlation. The correlation detected between two random variables A and B may not be, as it may well turn out that they are both correlated with some third CB C and the correlation we detected is only an echo of the deeper dependencies between A and C; B and C
In this sense, I am very attracted to the space of states in which we can collect many heterogeneous states and build a prediction on this population. The simplest ones apart from trend: trend acceleration, overbought/oversold, volume (not available), support/resistance.maybe something else.
The forum is full of developers of indicators and TS and I was hoping they would participate and I would enlarge the variety of models. However it hasn't happened yet. We are discussing an extremely primitive version of the model.
Non-stationarity is a property inherent in the world. Observation, on the other hand, is secondary, and depends on the tools and the way they are used.
In that sense, I'm very attracted to the space of states...
It seems to me that it is much worse than that. Management has considered deterministic processes, random processes and indeterminate processes. Deterministic and random processes have the property of jumping from one state to another (the philosophical transition from quantity to quality - Hegel). Indeterminate processes are processes with human involvement. If the control object is a process with participation of people, then it should be considered as an indeterminate process - a process which at different moments of time may behave as determinate (in formation, in step and with song), random (revolution) and their mixtures. It is based on the ability of people to organise themselves. The market is a typically indeterminate process: once a trend, once a sideways trend, once a panic.
Your reasoning is very sympathetic to me. Everything we see around us is not the result of exactly one cause. A person applying matstatistics should always keep this in mind. The basic concept of correlation. The correlation detected between two random variables A and B may not be, as it may well turn out that they are both correlated with some third CB C and the correlation we detected is only an echo of the deeper dependencies between A and C; B and C
In this sense, I am very attracted to the space of states in which we can collect many heterogeneous states and build a prediction on this population. The simplest ones apart from trend: trend acceleration, overbought/oversold, volume (not available), support/resistance.maybe something else.
The forum is full of developers of indicators and TS and I hoped that they will participate and I will add a variety of models. However it hasn't happened yet. We are discussing very primitive variant of the model.
What do you mean by "overbought/oversold"? Could it be the difference between the actual price value and its value according to the regression equation, i.e. its expected value?
So use it. Tell me exactly what is not clear to you? Where does the difficulty begin?
We have three types of variables:
Dependent - no problem, it is e.g. EURUSD
Independent variables: How much, just EURUSD, above it turns out to be better to take the dollar index, Not clear.
The state variables calculated from the independent variables and serve as arguments for the independent variable. How many and which ones? What real processes do they reflect? So far it is clear to me that we need to model trend + noise and then trend + noise in the residual from the first model. Maybe trend acceleration or something else?
What do you mean by "overbought/oversold"? Could it be the difference between the actual price value and its value according to the regression equation, i.e. its expected value?
Was the lack of a time range in the question originally intended? All processes tend towards an equilibrium state. Non-stationarity is the absence of detectable