Interesting data on the US stock market

tommcginnis · Oct 21, 2017

ironchef said:
Are there algorithm that can extract the trend from the variance?

You mean, like, where things regress??
http://stattrek.com/regression/residual-analysis.aspx?Tutorial=AP
Which means, you perform a regression, and then take a look at your data_vs_model.
These "residual" values -- leftover umph after the model was done -- often carry clues about underlying systematic behaviors. (And if it's systematic (non-random), it can *and*should* be modeled.)

So the answer to a question about how to extract trend from variance is our old friend "regression." But the *real* answer is in then plotting the residuals, and looking for non-random behavior.

[This just totally cracks me up.
I went to graduate school *nearly* innumerate. And the guy who schooled me on regression could've gotten good results (maybe better results) from a cinder block -- just a teaching *genius*. The thought that I would be referring back to lectures from >30 years ago ..... would certainly bring a smile.]

ironchef · Oct 22, 2017

tommcginnis said:
You mean, like, where things regress??
http://stattrek.com/regression/residual-analysis.aspx?Tutorial=AP
Which means, you perform a regression, and then take a look at your data_vs_model.
These "residual" values -- leftover umph after the model was done -- often carry clues about underlying systematic behaviors. (And if it's systematic (non-random), it can *and*should* be modeled.)

So the answer to a question about how to extract trend from variance is our old friend "regression." But the *real* answer is in then plotting the residuals, and looking for non-random behavior.

[This just totally cracks me up.
I went to graduate school *nearly* innumerate. And the guy who schooled me on regression could've gotten good results (maybe better results) from a cinder block -- just a teaching *genius*. The thought that I would be referring back to lectures from >30 years ago ..... would certainly bring a smile.]

What you are saying is after linear regression (trend extraction), I should look for second/third order effects and there maybe tradable information there.

Thank you for the insights Professor Tom.

Best to you.

tommcginnis · Oct 22, 2017

ironchef said:
What you are saying is after linear regression (trend extraction), I should look for second/third order effects and there maybe tradable information there.

Thank you for the insights Professor Tom.

Best to you.

"Bingo!" -- and nicely put.

As for the (over-the-top!) "Professor Tom" appellation, 'Yeah. In my dreams.'
But TOTALLY I just parrot whut I wuz taught.
"My guy" (Larry Bartels) was a dissertation-writing Ph.D. from Berkeley when he was teaching me at Rochester, and Larry worshiped *his* guy, Christopher Achen.
( this was my first taste of Sage pubs. Wow! And still cheap!)
Larry went on to Princeton (for the next million years), and it turns out, before he left for Vanderbilt, wrote this little beauty with Achen, who'd joined him there!!!:
https://press.princeton.edu/titles/10671.html
Now, just to get freaky -- did ya see the date of publication?? 2016??
So, long-story-short, these guys got more smarts and still enthuse about their work than can be believed. I be jus' a piker -- or better -- a parrot truly.

Phew.
Okay, time to flush the boiler/system in my 10,000 year old house, so's I can put the heat on next week, and not freeze (or blow up)..... Real world.