Regression equation - page 10

[hrenfx]

ivandurak:
If we calculate the variance as a slope of the regression line, then the small value of this dodgy variance will indicate the value of the kotir close to the straight line, imho it is a very good predictor. If we divide the angle of regression by the dodgy dispersion this indicator will show that the market is not efficient, prices go in one direction, to trade on the news searching for trends.

The linear regression says little about market efficiency. The topic of efficiency (correlations) is more or less covered in the Recycle approach of market correlations search. I have nothing more to say at the moment.

[Alexey Subbotin]

ivandurak:
If we calculate the variance as a slope of the regression line, then the small value of this dodgy variance will indicate the value of the quotient close to the straight line, imho it is a very good predictor. If we divide the angle of regression by the dodgy variance then this indicator will talk about the inefficiency of the market, prices go in one direction, to trade on the news looking for trends.

did that

[Alexey Subbotin]

Here, I dug on the Internet - a ready-made quantile regression on Matlab, the Frisch-Newton (polynomial) method.

Working.

[hrenfx]

alsu:

Here, I dug around on the Internet - a ready-made quantile regression on Matlab, the Frisch-Newton (polynomial) method.

No Matlab and not familiar with it. If possible, give examples of comparisons between MNC regression and quantile regression on BP price data (any sample with "bad" outliers).

[[Deleted]]

Cubic regression.

Three samples. Graphical representation.

[hrenfx]

j21:

Cubic regression.

Three samples. Graphical representation.

Now it remains to investigate the out-of-sample relevance of the constructed regression (by values before and by values after).

[[Deleted]]

Doesn't anyone else have any ideas?

[Alexey Subbotin]

There are ideas, it's the implementation that's the problem:))

[Alexey Subbotin]

I'll show you a picture soon, it's interesting - some of my assumptions are confirmed.

[Vladimir Gomonov]

alsu:
I'll show you a picture soon, it's interesting - some of my assumptions are confirmed.

I'm waiting for a picture. It's a very interesting topic. I read it with great interest.