Econometrics: let's discuss the CU balance sheet. - page 14

[Avals]

avtomat:

this is where the residuals have an even distribution. So what? A CU with this output should be thrown out just because the residuals are not normal? ;))))

if the residuals are even, it is not worth throwing out anywhere - it's a grail))

[Alexey Subbotin]

avtomat:

this is where the residuals have an even distribution. So what? A CU with this output should be thrown out just because the residuals are not normal? ;))))

The abnormality of the residuals is a problem of the regression model, not of the series.) Theoretically it is always possible to fit a model so that the residuals are normal, but is it always necessary? Normality is not directly related to stationarity, but its absence in many cases indicates that the wrong type (model) of trend is selected during detrending, i.e. all regression relationships are not taken into account.

[TheXpert]

Avals:
if the leftovers are even, you shouldn't throw them away anywhere - it's a grail))

The grail is in the grail even perhaps.

[[Deleted]]

TheXpert:
The grail is in the grail even probably.

this is where the residuals have a Cauchy distribution. So what? A CU with this output should be thrown out just because the residuals are not normal? ;))))

[TheXpert]

avtomat:

So what? A TC with this output should be thrown out, just because the leftovers are not normal? ;))))

You are now engaging in quackery.

[[Deleted]]

TheXpert:
You're quackery now.

;)))) not at all. I just want to make it clear that the normality of the residuals --- cannot be considered as a criterion in the evaluation of TC performance !

Here the residuals have a logistic distribution. So? A TS with such an output should be discarded, just because the residuals are not normal? ;))))

[[Deleted]]

And a counterexample: here the residuals have a normal distribution. So what? A TS with this output is preferable to the previous ones, just because the residuals are normal? ;))))

[СанСаныч Фоменко]

avtomat:

And a counterexample: here the residuals have a normal distribution. So what? A TS with this output is preferable to the previous ones, just because the residuals are normal? ;))))

As they write in criminal law textbooks: "a criminal case is fabricated by narrowing the evidence".

All the time I write - arbitrary variance changes are important. In a normal case, changes in the variance are not arbitrary, but what matters is the magnitude of those changes!!! and so it turns out that you fudged the pictures. A little nuance - the variance, and you're silent.

If you take the residual and calculate descriptive statistics, as I did above, you can see everything immediately in numbers, not by eye as in the pictures. Feel the difference.

[[Deleted]]

So keep playing with this proverbial "normality"... I understand how difficult it is to give up established perceptions... I wonder how soon you'll understand what I just said, and what I was trying to get across...

[Avals]

avtomat:

this is where the residuals have a Cauchy distribution. So what? A CU with this output should be thrown out just because the residuals are not normal? ;))))

just means not enough trades for Koshi to demonstrate himself)) And what does the residuals distribution look like? Not how you generated Cauchy, but what was the difference between trend and equity