Econometrics: one step ahead forecast - page 61
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
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?
That's not what I mean...
Which theoretical statements are unclear to you and cause problems?
Why don't you want to write your formula (18) as a gamma function from EViews?
You need to present the whole mechanism of calculating gamma-function parameters, I offered the existing exel variant, I ask again: is it suitable for you or not? If not, you can see in my branch how Sergeyev does it, it will be clear.
Was the lack of a time range in the question originally intended?
Here one record is not enough, you need to present the whole mechanism of calculation of Gamma-function parameters, I offered the existing exel variant, I ask again: is it suitable for you or not? If it's not, you can see in my branch how Sergeyev does it, it will become clear.
This is from TA, some qualitative state of the market. Overbought: volumes grow, the number of participants grows, but the price grows less and less, and then goes sideways
overbought/oversold is the use of the return property of real processes. There is the opposite - trending. Many sacred cows (as Matemat calls them) are nurtured by this property of trendiness - trend is your friend, cut losses let profits grow, etc. But real markets are multifaceted - they can be both returning and trending at different scales(timeframes) and time periods.
The trendiness/flatness can be formally assessed in different ways. And even in relation to some price derivatives, in particular regressions.
For this purpose, we can use the well-known Einstein's law, which can be the basis for separation of trendiness and flatness.
Let us take for example a close price as a reference point and analyze how the price moves away from it in time, and compare it with the way it does on the SB according to Einstein's formula. For any timeframe we have the following picture:
red is the case for the sb (deviation increases in direct proportion to the root of time), blue for prices (EURUSD h1, for other instruments and TF similarly). On the abscissa axis is time in bars, on the ordinate axis is how the standard deviation of the price changes. I.e. in relation to the last price the future prices change similarly to the random walk in terms of the variation of the deviation value.
Now let's take an exponential wave and similarly compare how the deviation of prices changes relative to it over time.
Let's take m1 EURUSD and three different dummies for comparison with periods of 12, 24, 60 (just to make them different :)). Here is the picture:
light blue random wandering. On the real data the cwm grows slower. As the Mach period increases the difference is more significant. That means reversion - prices tend to return to the wave.
Now let's compare with EURUSD h1. The wave periods are the same:
the picture has significantly changed. In comparison to the wave with a period of 12, prices tend to be trending. I.e., roughly speaking, if the price is higher than Ma(12) then we should buy, if it is lower then we should sell. In relation to Ma(24) neither trending nor flat (averagely), in relation to Ma(60) return.
For days:
trending is observed both relative to ma(12) and ma(24)
In general, it would be better for you to determine whether your regression is regressing or trending. From this depends on how to trade your model and whether it is worth trading at all. Of course this is a rough guessing level.
P.S. in your terms considered "prediction error"
Let's take for example the closing price as a benchmark and analyse how the price moves away from it over time and compare it with the way it does on the SB according to Einstein's formula. For any timeframe we have this picture:
And you can do the same but over the price retours and the chart is in double logarithmic scale.
Strange, EURUSD should not be equal to SB, the blue line should be above the red line.
overbought/oversold is the use of the return property of real processes
It is a reflection of the opinion of the crowd on incoming information. Overbought: something is too much of a good thing. and vice versa, too much of a bad thing. Retraceability has nothing to do with it. Trends have been around for years.
I don't understand anything. Are there any publications on the subject.
Now in order.
How is a 1-step-ahead forecast made with a dashboard? The far right value of the waving is calculated after the price arrives. I know this problem well. That's why in my model the far right value is calculated based on the 4 previous ones. The +1 value is calculated from the 4 available measured values. But what about you?
Let's deal with that and then the rest.
Strange, EURUSD should not be equal to SB, the blue line should be above the red line.
Why?
Let us now take an exponential scale and compare the deviation of prices relative to it over time in a similar fashion.
Let's take the m1 EURUSD and three different charts for comparison with periods of 12,24,60 (just to be different :)). We have the following picture:
The light blue one is a random wandering. On the real data the cwm grows slower. As the Mach period increases the difference is more significant. This means reversion - prices tend to return to the mask.