Statistics as a way of looking into the future! - page 3

 
m_a_sim писал(а) >>

The forecast function is written as f=T*S, where T is the trend component (regression) and S is the seasonal component. The seasonal component is calculated in a certain way, and it has 24 values. I didn't make it up, it's written in the book. The prediction is supposed to be one hour ahead, in the picture it is 121 hour and then it is missing the required values X1 and X2, these values are calculated with the Prediction function in EXCEL, which means you better trust the prediction at 121 hour and after that there may be some differences

Prediction process must necessarily include calculation of possible errors (error ellipsoid).

See fig. If we are at point 0(t=0) and we predict price C2, then possible errors in price and time must be calculated. Then we can calculate the probability of hitting this area, otherwise as mathematicians say, the probability of hitting the point = zero.

 
Prival >> :

The prediction process must necessarily include calculation of possible errors (error ellipsoid).

See figure. If we are at point 0(t=0), and we predict price C2, then the possible error in price and time must be calculated. Then we can calculate the probability of hitting this area, otherwise as mathematicians say, the probability of hitting the point = zero.

Something I don't understand.... Is the process of finding the probability itself important?

The model can be estimated in many ways, e.g. average absolute error.

Can you tell me how to find the probability of hitting a point?

 
Determine the distribution law of the errors generated by your model. Use this to calculate the probability of their values falling within the quantile you want.
 

to m_a_sim

I guess my request to show the incremental cloud (see my post above) went unnoticed.

I would still like to see a picture.

Prival писал(а) >>

The prediction process must necessarily include calculation of possible errors (error ellipsoid).

See fig. If we are at point 0(t=0) and predict the price of C2, we must calculate the possible error in price and time. Then we can calculate the probability of hitting this area, otherwise as mathematicians say, the probability of hitting the point = zero.

Prival, what is the deeper meaning of considering two values in the model - price and time? Wouldn't it be easier to limit it to one, or the model will suffer from it. Maybe there are some principled considerations on this point?

 
Neutron >> :

to m_a_sim

I guess my request to show the incremental cloud (see my post above) went unnoticed.

However, I would like to see the picture.

Prival, what is the deeper meaning of considering two values in the model - price and time? Isn't it easier to limit yourself to one, or the model will suffer from it? Maybe there are some principal thoughts on it?

I'm a little confused about what to plot from what, let's use my definitions, t-time, X1 and X2-factors, f-factor of forecast, y-regression

 
m_a_sim писал(а) >>

I'm a bit confused about what to build from what, let's use my definitions, t-time, X1 and X2-factors, f-fonction of the forecast, y-regression

You have two time series - the original price series (black) and the result of your forecast (red):

Now build for each of them a series of first differences (increments, Returns) and plot one on the abscissa axis, the other on the ordinate axis.

That's it.

 
Neutron >> :

You have two time series - the original price series (black) and the result of your forecast (red):

Now construct for each of them a series of first differences (increments, Returns) and plot one on the abscissa axis, the other on the ordinate axis.

That's it.

 

Draw a straight line through this cloud using the method of least squares and see the tangent of its slope. So, at first glance, the result is good! But we need figures. As I understand it, the points are plotted on the axes.

For which instrument is the forecast and on which TF?

 
Neutron писал (а) >>
Draw a straight line through this cloud using the method of least squares and see the tangent of its slope. And so, at first glance, the result is Good! But we need figures.

Mm. And where is it good, if you ideally want a straight line at 45g?


P.S. For example on the original chart I already see the trading fatal error in the training period. What to say about the forecast then?

 
Neutron >> :
Draw a straight line through this cloud using the method of least squares and see the tangent of its slope. So, at first glance, the result is good! But we need figures.

tg=0.3945 angle 22 degrees