Obtaining a stationary BP from a price BP - page 2
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
I wonder if anyone has ever had to get white noise in price conversions?
On the doji.Nerd bluster. Isn't your own brains enough to realise that everything in the link you cited is nonsense?
Read on, and I quote: "Limitations. Recall that the ARPSS model is only suitable for series that are stationary (mean, variance and autocorrelation are approximately constant over time); for non-stationary series, take differences. It is recommended to have at least 50 observations in the source data file. It is also assumed that the model parameters are constant, i.e. do not change over time. " (I don't want to discuss the figure of 50 observations, because it is clear even to a fool on this forum that 50 transactions is not a result)
Suppose we have a non-stationary series, we took the residuals - delta(x). The residuals themselves, as assumed in this nerdy "work" should meet the requirements, and I quote: "containing only noise without systematic components".
Fuck it. Let there be noise. The noise itself is not predictable in any way. Therefore, it is useless to approximate it. But it does have the property, and I quote: "The residuals shall be normally distributed and have MO=0."
Hence, instead of noise we take its MO=0.
Substitute it into the forecast: forecast(time + i + j) = price_appr(time + i + j) + delta_appr(time + i + j) = price_appr(time + i + j) + 0 = price_appr(time + i + j)
So, the forecast on noise is the first approximation: price_appr(x). And the first approximation, as I said in the third post of this thread, is a naked fit. The result is:
Botanical prediction = fitting
What's all the fuss about? Take ZigZaz and basta, that tail that twitches - wait till it stops twitching. And seriously: this new stationary series model may represent the original non-stationary one. Where they discuss ARPSS they also discuss confidence intervals between the original BP and its model. I can't tell where the botanists and the zoologists are.
I wonder if anyone has ever had to get white noise in price conversions?
In its pure form, no one. White noise has equal amplitudes in all harmonics from 0 to infinite. It is not found in nature in its pure form, because there are no such ideal acoustic conditions in which any harmonics are not amenable.
To check for white noise, you can take the first N harmonics and compare their amplitudes. If they are roughly the same, then the BP is noisy.
Ограничения. Следует напомнить, что модель АРПСС является подходящей только для рядов, которые являются стационарными (среднее, дисперсия и автокорреляция примерно постоянны во времени); для нестационарных рядов следует брать разности. Рекомендуется иметь, как минимум, 50 наблюдений в файле исходных данных. Также предполагается, что параметры модели постоянны, т.е. не меняются во времени.
The author of the paper wrote some nonsense, because the mentioned model, and progenitor models, can be perfectly well used for non-stationary series (the coefficients will just be non-stationary). Significant prediction errors lie elsewhere - in a meaningful mismatch between the source distribution and the model used. In other words, the necessary condition is that the distributions of the ARPSS and the price series coincide, which of course is not the case in nature.
PS: by the way, some glitch, quote selection doesn't work Hmmm, quote selection works, but separate from text selection (IE7),
Nerd bluster. Isn't your own brains enough to realise that everything in the link you cited is nonsense?
Read on, and I quote: "Limitations. Recall that the ARPSS model is only suitable for series that are stationary (mean, variance and autocorrelation are approximately constant over time); for non-stationary series, take differences. It is recommended to have at least 50 observations in the source data file. It is also assumed that the model parameters are constant, i.e. do not change over time. " (I don't want to discuss the figure of 50 observations, because it is clear even to a fool on this forum that 50 transactions is not a result)
Suppose we have a non-stationary series, we took the residuals - delta(x). The residuals themselves, as suggested in this nerdy "work" should meet the requirements, and I quote: "containing only noise without systematic components".
Fuck it. Let there be noise. The noise itself cannot be predicted in any way. Therefore, it is useless to approximate. But it does have the property, and I quote: "The residuals shall be normally distributed and have MO=0."
Hence, instead of noise we take its MO=0.
Substitute it into the forecast: forecast(time + i + j) = price_appr(time + i + j) + delta_appr(time + i + j) = price_appr(time + i + j) + 0 = price_appr(time + i + j)
So, the forecast on noise is the first approximation: price_appr(x). And the first approximation, as I said in the third post of this thread, is a naked fit. The result is:
Botanical prediction = fitting.
This is a test of the adequacy of the prediction model. Residuals can be taken not only of first order. I.e. delta2(time + i) = Open[time + i] - forecast(time + i). The method only says that the prediction model is adequate. In your case, the prediction model is the forecast. I.e.
"4. Checking delta(x) for white noise. If it's noisy, bummer grandma. If it doesn't make noise, keep going."
it's not bum bum, it's the opposite - the prediction model is good. The residuals have no systematic component, are independent. As long as this is not the case, you can build models extrapolating the residuals to infinity. It is a criterion of stopping that we have come to what we are looking for.
Is it hard to figure it out yourself? ;)
And to check for white noise, you can take the first N harmonics and compare their amplitudes. If they are roughly the same, then the BP is noisy.
Wouldn't you rather calculate the ACF?
This is a test of the adequacy of the prediction model. Residuals can be taken not only of first order. I.e. delta2(time + i) = Open[time + i] - forecast(time + i).
Again it is nonsense.
As a result we get: delta2(x) = delta(x), because, forecast(x) = price_appr(x).
How hard can it be for you to figure it out?
Wouldn't it be better to calculate the ACF?
Why don't we pay attention to the tail in ZZ? It's a prediction for sure. And you still have to prove that you have it.
Again, this is nonsense.
The result is: delta2(x) = delta(x), because, forecast(x) = price_appr(x).
Is it so hard to figure it out yourself?
Do a little more thinking :) Just because the MO of a random variable = 0 does not mean that the CB itself can be replaced by zero, as you cleverly do :) :)