Econometrics: one step ahead forecast - page 123
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
Cleaned up swearing, flooding and off-topic posts. Totals (by deleted posts):
These are facts, not emotions. Please draw your own conclusions. All deleted posts are documented.
It's just the facts. Really, I'm just being an angel here - but I didn't mean to, honestly.
great fuckin', aaaa...... come on.
Goodbye and goodbye! I wish you good luck in other forums!
Uh... What was the quote again? I think I read it, but it was long ago.
that was for the troll, I'll remind you more:
- What is it? - asked Woland in bewilderment and began to look at the board, where the officer standing on the king's cage was turning away and covering himself with his hand.
- You wretch," said Woland thoughtfully.
- "Sire, I appeal to logic again," said the cat, pressing his paws to his chest, "if a player announces check to the king, and meanwhile the king isn't on the board, check is declared invalid.
- Do you give up or not? - Woland shouted in a frightening voice.
- Let me think," the cat answered meekly, put his elbows on the table, put his ears to his paws and began to think. He thought long and hard, and at last he said: "I give up.
- Kill the stubborn creature, Azazello whispered.
– Убить упрямую тварь – шепнул Азазелло
I sprinkle ashes on my head.
As far as I know, in certain places provocateurs and their accomplices are punched straight in the face, and not listened to about their angelic nature
Here's the result.
Taking from the beginning of the sample
Take starting from 100 bar
Take the trend change
Take the sideways
All different.
We take D1
It doesn't look like H4 at all.
How different are they really?
In case ACF really characterizes the memory length of the market(it's written in the books, but not the fact at all!), ACF value on D1 lag 1 should approximately correspond to the mean ACF value of lag 1 to 6 for timeframe H4, and D1 lag 2 corresponds to the mean lags 7 to 13 and so on, but this is not happening and ACF D1 is still closer to the value of the same lags on H4 ... In addition, if you take a sample of data over a longer period of time, the ACF coefficients of different timeframes of the same market will differ less and less...
How dissimilar are they really?
In case ACF really describes the memory length of the market (it's written in the books, but not the fact at all!), then the ACF value on D1 lag 1 should approximately correspond to the mean ACF value of lag 1 to 6 for timeframe H4, and D1 lag 2 corresponds to the mean lag 7 to 13 and so on, but this is not happening and ACF D1 is still closer to the value of the same lags on H4 ... In addition, if you take a sample of data over a longer period of time, the ACF coefficients of different timeframes of the same market will differ less and less...
Still, your approach reeks of "all or nothing".
Regressions are the simplest tool. Used to demonstrate the problem of predictability. Even regressions proved inaccessible to most.
The approach itself was different. Pinch off a piece of the unexplored and unknowable, but systematically. Fractals are more of an experiment and there are many issues of prediction accuracy and confidence behind the scenes. That's why EViews, which gives systematicity and isn't very restrictive in the set of methods. If you take into account that it has output to Matlab, the limitation is its own brains.
Therefore. Trying to solve problems piecemeal.Also, I don't really understand how you can systematically pluck from the unexplored.
Look, I can't stand arrogance on a level playing field. What you're really demonstrating, I'll keep quiet.
Oh, shit, first of all it's not so, you have to understand the subject. Secondly - do you really think that by applying enwil you get reliable estimates?
EViews only does model identification and model prediction. All this is in matlab, statistics, mathematics, spc, etc. But, none of these models are true. You're just fooling envil by slipping false correlations
Everyone has their own Dao. :о)
Also, it's not very clear to me how you can systematically pluck from the unexplored
All science is just that: solves what it sees, and from what it sees, what it can, and from what it can, what it can apply.
Look, I can't stand arrogance on a level playing field.
Arrogance has nothing to do with it. The simplest model was taken to demonstrate a different problem and a more important one - predictability. Everyone has focused on regression, with many posts with the usual nonsense, people don't know the terminology and don't take it personally that doesn't apply to you.
Do you think that by applying enwil you get reliable estimates?
EViews just does model identification and model prediction. All this is in matlab, statistics, mathematics, spss, etc. You are just fooling envil by slipping false correlations
You have to specify the stovepipe before you can evaluate anything. And this is TA, compared to which EViews is a huge step forward.
Statistics teaches you not to trust anything at all, including estimates. But a systematic calculation of estimates for any model is a step forward
But, after all, none of these models are true.
My descriptive model: cotier= trend + noise + periodicity + seasonality + outliers.
I'm starting to dip sequentially on what I can, and I can: trend + noise. Big forecast with TA again - it doesn't recognise noise at all. I know the reason for bad forecast result, I don't see any reason to post it due to low public interest.
What else can I add to my model? It's not complete, but a constructive one is needed.Briefly, in abstracts:
(1) this:
котри= тренд+ шум+ периодичность+ сезонность+ выбросы.
impossible to find, moreover, it is not true. there is no seasonality, no periodicity, no noise(!!!). This model does not work. Among other things, you will not be able to work with a quotient directly. I strongly recommend looking for a transformation that brings kotir to at least some stationarity
(2)
Everyone has focused on regression, with many posts with the usual nonsense, people don't know the terminology and don't take things personally that don't apply to you.
This is me trying to concentrate you on more important things. It is clear that not everyone is a mathematician etc.
(3)
You have to specify the stovepipe before you can evaluate anything. And this is TA, compared to which EViews is a huge step forward.
TA is not a "stovepipe", it's utter nonsense, though - colleagues who have been here a long time know my dislike of the thing.
(4)
The problem and the more important one - predictability.
By the way, how do you assess predictability?
Exclusively intuitively as the probability of a prediction being fulfilled.