Remembering veterans: Box and Jenkins - page 10
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
Probably more appropriate: price - f1 + f2? Because in your case the trade signal must be triggered at the level of the sum of the two scales, i.e. the price must cross the level at about two prices in order for the signal to change its sign.
If it is more correct to assume that the residuals between the TS and the quote for the case of two masks crossing, where: price is the current price, f1 and f2 are the last readings of two muvings with different periods and the formula: residuals = price - f1 + f2, then we obtain that the residual equals equity, assuming that the spread is zero.
And if that's true, then your statement, and I quote:
You can only trust the test and forward test if the residuals = kotir - the TS is stationary! i.e. passes the unit root test.
is absurd. Absurd because:
1. Whether there are fixed equities for CU, I don't know, as I've never seen any fixed equities.
2. Stationarity implies lack of trendiness, i.e. such a TP is a perfect sideways move. If equity is stationary and stationary BPs have MO = Const, this very MO will be around the initial deposit.
So, don't engage in epigonism, but learn the maths - it rules. And give up worshipping the frauds, such as Boxes and Jenkinsons - sectarianism has not brought anyone good, except the founders of sects. However, some founders didn't end well either.
I am not discussing the stationarity of the TC. Instead of idiotic advice, sit down and spend a week on a book at specialist rather than bachelor level, and then we will discuss. It makes sense to discuss when the opponent does not confuse TC with the remainder.
... May I ask, what are you condemning? So that the average ktn would understand it as well?
Apparently, something that doesn't make sense. Otherwise what's the point of discussing something that is trivial even to a drunken hedgehog?
...
In the following thread I will give the results of the calculations for this model. I suggest to discuss the results and their applicability to Forex.
...
faa1947:
...
At least one conclusion: indicators without adaptation to the current quote are meaningless.
There's a lot to make up.
We are talking about very specific things that I understand and understand equally with millions of other people.
What you have described is still a numbers game. Maybe correct, maybe not. You have not provided any evidence. Generally accepted deductions on stationarity and non-stationarity I have cited. I reply that my calculations are in line with the calculations people have been using for 40 years.
Once again, just in case - where are the successful traders who understand as you do? You do not have to name a million, at least one or two.
What I have described is an abstract example that shows that
1. any statistics on the entire range is the mean temperature in a hospital. Including non-stationarity test
2. It is not necessary to reduce all non-stationary series to stationary ones in order to earn money. The distributions of deals i.e. sectors of the chart between entry and exit points as well as series of deals are stationary. In addition to stationarity there is a positive EA of course. Otherwise the stationarity is of no use.
Stationary is the distribution of trades, i.e. sections of the chart between the entry and exit points, as well as the series of trades.
There is no point in exhorting, because Mr. Econometrician is unwilling to discuss the stationarity of the TS.
Read the scriptures of Box and Jenkinson and don't twist what econometricians have atrophied.
I will summarise how I understand Box-Jenkins and their practical refraction.
Going beyond the 1972 book mentioned above, the scheme is as follows.
1. we take a cotier.
2. With the unit root test we check this quotient for stationarity.
The upper table shows Prob = 0.7350, which does not allow rejecting the hypothesis of non-stationarity of the quotient. It is convenient to assume that the probability of non-stationarity = 75%.
3. Let's differentiate the quotient according to Box and Jenkins - take the difference between the neighboring bars. We obtain a differentiated series, i.e. series I(1) in the ARI(1)MA model
4.check this new series by unit root test
We see that this series is stationary (weakly or covariantly stationary, i.e. the mean and autocovariance of the series do not depend on time). Prob = 0, i.e. we strictly reject the hypothesis of non-stationarity of the series.
5. For the incremental series we write down the ARMA equation:
d_eurusd ar(1) ma(1) c @trend
6. We estimate and get the result:
We see that we cannot reject the hypothesis that the coefficients at C and @TREND are equal to zero. That is why we shall exclude these two independent variables from the model.
This is the final result.
The last two lines of the table are the most valuable: as the values are quite far from 1, this indicates the stability of the resulting model.
7. What have you got? We got an indicator on the basis of which we can build our TS. In this case we can trust the results of testing in MQL.
What is the merit of Box and Jenkins?
They wrote their book at a time when a vastly efficient market was randomly wandering through a future Nobiles. Under these circumstances they say: the market is unsteady and there are dependencies between neighbouring bars. Their ARIMA model is still used in economics, some economic bureaus of the US government have put out in the public domain the model by which they do their calculations.
A lot has been done in 40 years, which has limited the randomly wandering noobiles. In my opinion the most significant two things have been done.
1. That the above managed by differentiation to get a stationary residual (which allowed ARMA to be applied) is not at all necessary. The residual can still be non-stationary. The ARCH model was developed for these cases.
2. It is not clear how to apply ARMA in multicurrency cases. It is not clear how to include a variable such as "market sentiment" in the model. We use "state space" models which are much more powerful, flexible and effective in the fight against market instability.
That's all for now. Everyone is free to correct the mistakes, and to supplement the incompleteness.
Thanks to everyone who has spent time on this topic.
I carefully read the whole article and the comments, and I fully support Mathemat in explaining things to people in a much simpler way: Mathematics taught at university is best not used in the market - it will not make money in your pocket, but that does not mean that mathematics is bad. OK, now closer to the subject of Box & Jenkins. Their entire book is devoted to the analysis of time series, which in modern literature are classified as DS (difference stationary) - series that become maybe "weakly stationary" after taking the difference(s). Then a simple statistical model is constructed for the series, which is as far away from the financial series as the Earth is from Mars. It can be easily proven empirically, but unfortunately you can't check the quality of the model with packages like Statistica and Eviews, it can only be checked with your own account, and if there is extra money in your pocket, the model must have gotten closer, like the distance between Earth and the Moon. From the whole book we should make a very simple conclusion that a series of quotes is non-stationary and the quotes slightly depend on the past values - the models ARMA(p,q), ARIMA(p,d,q) where the parameter d is an integer, FARIMA(p,d,q) where d is a fractional number, and if you dig deeper and learn that the quotes also depend on the lag values of squared price logarithms and that there are also empirical and theoretical studies which support the conclusion that the volatility of the logarithm of price movements of financial assets is a time-dependent stochastic process like ARCH or GARCH, then of course must reread this book again.
And if you want to hit the wall in general, I would advise to read this book:
1)К. Granger, M. Hatanaka "Spectral analysis of time series in economics".
2)Е. P. Churakov "Forecasting Econometric Time Series".
3) J.S.Bendat A.J.Pearlson "Measurement and Analysis of Random Processes" Mir, 1967.
4)J.S.Bendat A.J.Pearson "Applied Analysis of Random Data" Mir, 1989
5)R.N. Mantegna, G.J. Stanley "Introduction to Econophysics. Correlations and complexity in finance".
P.S. I am convinced that if you study these 5 books thoroughly and do not trade without MM, your account will not suffer much during this time.
I gave the example of the mashups above. In EViews it is the button that calculates this residual.
I will give you the picture again.
So.... The picture is the one?
So.... The picture that?