Econometrics: one step ahead forecast - page 10

New comment
 
faa1947:
One last time: the possibility of a step-by-step assessment of the quality of the development. And we don't take the other one, we refine it. And not mine at all - it is used by millions of traders around the world.

there is step-by-step model fitting to get HP residuals, which means nothing in terms of profitability of the strategy. It's a fit, albeit a scientific one
 
gpwr:


Will we open a position on every bar? If not, under what conditions? If a position is opened when the prediction exceeds two spreads, then extrapolation errors should only be calculated for those conditions. And you want to know the extrapolation errors regardless of entry and exit conditions. Indeed, I am wasting my time here.


A random variable and its error are inseparable concepts. You don't understand what's being written here
 
Avals:
There is a step-by-step model fit to get the NR residuals, which means nothing in terms of profitability of the strategy. It's a fit, albeit a scientific one.

Look at the model (formula) at the beginning of the topic. There is an HP indicator and the difference (error) between the indicator and the quotient. Then there is the process of estimating the coefficients of the equation by OLS, i.e. so that the error (mismatch between the quotient and the model) is the smallest. Not out of the blue, but the smallest. It's a long way from being profitable. But why talk about model profitability if you don't know if the model matches the quoted price?

And be careful about being scientific, especially when you call internationally recognised science scientific.

 
faa1947:

Look at the model (formula) at the beginning of the topic. There is an HP indicator and the difference (error) between the indicator and the quotient. Then there is the process of estimating the coefficients of the equation by OLS, i.e. so that the error (mismatch between the cotier and the model) is the smallest. Not out of the blue, but the smallest. It's a long way from being profitable. But why talk about model profitability if you don't know the fit of the model to the quotient?

What guarantees that the model is consistent with the quoted price - does it pass all your tests?
 
Avals:
what ensures that the model meets the quote - passing all your tests?
The simplest answer is R-squared, but this is under conditions and therefore not true. At this stage: R-squared and preferably a strict rejection of the hypothesis that the model coefficients are equal to zero. This is just the beginning of the journey. If the residual has AK, then the resulting correspondence between the model and the quotient cannot be trusted. In general, the residual must be stationary. In our first step, this probably cannot be obtained.
 
P.S. The indicator does not predict anything at all. It is the rules built using it that make predictions. And not necessarily only on it. And not necessarily the buy or sell signals on every bar. How to analyze the indicators, etc. step by step?
 
Avals:
P.S. the indicator does not predict anything at all. It is the rules that are built using it that makes predictions. And not necessarily only on it. And not necessarily the buy or sell signals on every bar. How to analyze the indicators, etc. step by step?

I totally agree. Extrapolating an indicator is nothing more than extrapolating a formula. There is no randomness.

By the way, I remembered a branch opened by Reshetov. He transformed a non-stationary series into a stationary one. A related branch, but without EViews systematisation.

 
TheXpert:
Are you going to continue your lectures on neurons?

I will. There's one lecture left. Work got a little hectic. Maybe I'll post it this weekend.
 
faa1947:

A related branch, but without EViews systematisation.



I forgot to ask... why EViews and not gretl for example...? http://gretl.sourceforge.net/ Russian and free... The developers' motto is "From econometricians, for econometricians".
 
faa1947: One last time: the possibility of a step-by-step assessment of the quality of development. And we don't take the other one, but refine it. And not mine at all - millions of traders in the world use it.

What the hell millions, my colleague. God forbid a couple of thousand of them understand what they are doing...

(Apologies if I'm a little late in responding: the topic is growing too fast. I'll respond as I read the comments. And in general I like that the topic is so dynamic...)

1...34567891011121314151617...139
New comment