The Sultonov Regression Model (SRM) - claiming to be a mathematical model of the market. - page 10

 
Avals:
the prediction model is correct if the residuals are normally distributed.
Um, so it turns out that a model without residuals is a deterministic component?
 
Demi:


If the chosen regression model describes the true relationship well, then the residuals should be independent normally distributed random variables with zero mean, and there should be no trend in their values.

What kind of stationary is there?

P.S. Been to the mammoths, come back, back with you...


Yes, of course. But the remainder is tested by the unit root test, which is stationarity.

Another problem. What if it's not exactly as you wrote? And if it's like you wrote, can we trust the prognosis?

 
TheXpert:
Um, so it turns out that the model without residuals is a deterministic component?

it means that the variables are deterministic and not random
 
faa1947:


Yes, of course. But the residual is tested by the unit root test, which is stationarity.

Another problem. What if it is not exactly as you have written? And if it is as you have written, can the prognosis be trusted?


If the input variables are normally distributed, stationary, the model residuals are normally distributed and the prediction accuracy R or R2 is satisfactory - we can! And we need to!
 
TheXpert:
Um, so it turns out that the model without residuals is a deterministic component?

A model without residuals is a model that predicts series values without error. The residuals are the error (the difference between the predicted value and the real value). So there is actually a decomposition into a deterministic component (forecast model) + noise (normally distributed residuals)
 
yosuf:
I do not understand what you mean by discrete series prediction? The result of processing the data presented is that the discrete series has MO = 0.878649833 and is significantly skewed towards 1. Am I still supposed to determine the predictive alternation of ones and/or zeros? An absurd requirement when dealing with discrete series. I'm sure if you somehow calculate the sum of this series and divide by the number of "throws" you will get the above result.


That series contains 45 zeros and 45 units. Expectancy = 0.5.
 
faa1947:

18 is an analytical formula. Calculate the function value from it and take the difference from the quotient. We get the smoothing error. Let us start working with this error. Or did I miss something?
You can try it. Here is an indicator that implements (18), maybe programmers will be able to implement this operation?
Files:
 
Demi:

If the input variables are normally distributed, stationary, the model residuals are normally distributed and the prediction accuracy R or R2 is satisfactory - we can! And we need to!

It doesn't happen in the market. Cotier is non-stationary, and the definition of non-stationarity that we use is too narrow for a real series.
 
anonymous:

That row contains 45 zeros and 45 ones. The expectation is 0.5.
How do you explain the fact that RMS has raised the ME to 0.8787? Moreover, if RMS input strictly alternating 0 and 1, then it also shows 0.5. So, there is a circumstance in the series you have given which shifts this equilibrium towards 1.
 
faa1947:

There is no such thing as a market. Kotier is non-stationary, and the definition of non-stationarity we use is too narrow for a real series.

Well, then a regression model would be a dead giveaway. There are a lot of specialists who know regression analysis, but only a few make money in the market.