What makes an unsteady graph unsteady or why oil is oil? - page 17
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
On the topic, i.e. what makes price movements a non-stationary random process,
you must first define what non-stationarity itself is.
Non-stationarity is the change (not constancy) of mathematical expectation and dispersion
in time, i.e. the mean-dispersion on M5 derived from the last, say, 100 values is not
is not equal to the average-dispersion obtained at the same timeframe 24 hours ago.
on graphs in 365, which were cited faa1947, this very non-stationarity can be clearly seen
on the first graph - a Markovian, autoregressive process, on the second one it seems to be mixed
autoregressive-slidingaverage and both processes as I said are non-stationary and
you have to predict it first of all by bringing it to a stationary one.
For example take differences, i.e.: Difference(2)=price(2)-price(1)..... Difference(100)=price(100)-price(99),
and then apply the appropriate linear prediction model
Ой как интересно. Что технически означают термины "убегать" и "ловля убежавших"?
If you place a large lot with a limit.order so that it is visible in the stack, then in a thin market the prices of the nearest bid and ask - the centre of the stack - begin to move away from it. Sometimes this sets off a chain reaction - the "axe" moves out of sight of the cup. Then you hit those prices practically on the market for a while, and then you take your "axe" off. The market goes back practically to where it was.
I myself have dabbled (on the rayke, among others), until I was hit on the head once - my bid (someone from a large company) was satisfied, very (!))) partially satisfied his, the market went in the opposite direction, and then these buggers withdrew their unsatisfied bid - that is, they played this game. Thank goodness, the market did not move much at the time and my losses were minimal.
I take it that the first is to move their bids away from the axe. And the second is to increase volatility in the vicinity of the axe bid?
as I said, it is non-stationary and
You have to predict it first of all by bringing it to a stationary one.
For example take the differences, i.e.: Difference(2)=price(2)-price(1)..... Difference(100)=price(100)-price(99),
and then select the appropriate linear prediction model
IMHO this is a simplification of the problem. Smaller issues aside: depending on the parameters of the ARPSS model there are several varieties of models and it is argued that by identifying on the past model it is possible to make predictions for the future. To me it is obvious that this can be done if the ARPSS model itself has not changed in future periods. I found no such assumption in Box. So the problem of prediction for non-stationary models remains in the sense I mentioned.
ИМХО это упрощение проблемы. Если отбросить более мелкие вопросы, то: в зависимости от параметров модели АРПСС существует несколько разновидностей моделей и утвердается, что идентифицировав на прошлом модель можно сделать прогноз на будущее. Для меня очевидно, что это можно сделать, если не поменялась сама модель АРПСС на будущих периодах. У Бокса я не нашел такого предположения. Так что проблема прогнозирования для нестационарных моделей остается в указанном мною смысле.
"the ARPSS model itself" is exactly what the Auto-Regression-Integrated Moving Average (ARMIA) model is,
I understand auto-regression and moving average, but what do you mean by "pro-integrated"?
as in "the problem of forecasting for non-stationary models remains in the sense I mentioned."
Maybe Box's direct instructions to bring non-stationarity to a stationary form on the subject of prediction
are those. "assumptions" that is? :)
but whatever.
"the ARPSS model itself" is exactly what the Auto-Regression-Integrated Moving Average (ARMIA) model is,
I understand auto-regression and moving average, but what do you mean by "pro-integrated"?
in relation to "the prediction problem for non-stationary models remains in the sense I mentioned. "
Maybe Box's direct instructions to bring non-stationarity to a stationary form on the subject of prediction
are those. "assumptions" that is? :)
but whatever.
With Box, "reintegrated" is differences of an order higher than one, i.e. differences of differences until we get a stationary process. Box produces a stationary form, but for historical data, what will happen for future data? Will the ARPSS model there have the same parameters as for the historical data. This is what was meant.
1. Ну дык этот алгоритм Ваш или чей? Да и тут и мудрствовать много не надо ибо пипсовку на М1 легко распознать по коротким TP.
2. Ну дык хто ж Ваш алгоритм знает шоб шото помыслить? Ну а Ваше предыдущее описание работы по "порывам" выглядит извините не очень серьезно шоб даже экспериментировать с этим.
Видимо вам доставляет огромное удовольствие показывать собственную глупость... Так как если из всего выше написаного вы не поняли что написал Urain, то что вы тут вообще делаете? Просто достали ваши дурацкие коментарии типа "я все знаю а ты лол"... Только захламляют ветку и отвлекают от темы... Угомонитесь уже... Все уже все поняли про вас...
Can you say something about what I said without getting personal?
Please note that I am not making a personal reference to you. You think I wouldn't?
А по сути моих высказываний шото можете аргументировано заявить без перехода на личности?
Прошу заметить шо я на Вашу личность не перехожу. Думаете не мог бы?
If there was a point in your statements, I lost it a long time ago... The remaining impression of you in this thread: Do not understand what we are talking about (or pretend not to understand), but trying to prove that the man is wrong and generally a moron, as he "uses" m1 for prediction. And I was not getting personal, I just asked you not to spoil a branch, as a branch has given an interesting idea, and your comments are strongly affecting the idea of really useful ideas ...
With that I take my leave and do not want to join you in spoiling this thread... Have a nice day...
Если в ващих высказываниях и была суть то я ее давно потерял... Оставшееся впечатление о вас в этой ветке: Не пониаете о чем речь (или делаете вид что не понимаете), но пытаетесь доказать что человек не прав и вообще дебил, так как "использует" м1 для прогноза. Да и я не переходил на личности, я просто попросил вас не портить ветку, так как ветка подала интересную идею а ваши коментарии сильно мешают вникать в суть высказывания действительно полезных идей...
За сим откланяюсь, не хочу присоединяться к вам в деле обессмысливания ветки... Всего хорошего...
Dear Sir, you have only been asked to speak up and make your arguments, haven't you? Although your behaviour is quite understandable.
If you didn't understand the point of your statements or even lost it, I had nothing to do with it. But I can offer you a few ideas so that your neurosis doesn't get to you completely, which would be a shame. In any case I wish you good luck and recovery from all your illnesses and disorders!
Любопытно прогнатьть разные таймфреймы через анализатор. Это EURUSD M1 02/05/20000 по 01/06/2000 - всего 480 часов.
а это теже 480 часов но на на таймфреме Н1
Эту работу моожно было бы и не делать: временные ряды на М1 и на Н1 - это разные временные ряды с разными характеристиками. А то, что один получен из другого ни о чем (для меня) не говорит.
Of course they are. Because you have made them so. The way the older timeframe data is derived produces spectral components that are not in the original series.
What is the point of analysing something that does not exist?
Besides, if this is a Finvar analyzer, I have big doubts about its quality. They can't even calculate simple filters correctly.