Discussion of article "Exploring Seasonal Patterns of Financial Time Series with Boxplot" - page 31
You are missing trading opportunities:
- Free trading apps
- Over 8,000 signals for copying
- Economic news for exploring financial markets
Registration
Log in
You agree to website policy and terms of use
If you do not have an account, please register
the fact is that it's not always large for overlapping samples.
Theoretically, the correlation for such intervals can even be zero in some cases.
Theoretically, the correlation for such intervals can even be zero in some cases.
It is. You can still compare the intervals to each other, right? Or is that rubbish?
that's the way it is. It's still possible to compare intervals with each other, right? Or is it rubbish
For data of any nature, high correlation for overlapping intervals mathematically follows.
Based on that, one can decide for oneself if it is worthwhile.
For data of any nature, high correlation for overlapping intervals mathematically follows.
Based on this, you can decide for yourself whether it is worthwhile.
zero correlation can be considered high correlation? I'm confused, gentlemen.
A zero correlation is a high correlation? You're confusing me, gentlemen.
The average correlation will be high. Sometimes some cases will be near zero.
The average correlation will be high. Occasionally some cases will be near zero.
but is the relative difference significant? between a pair of clocks with an average correlation of 1 and another pair with 0.
The permutation approach doesn't work (non-intersectional sampling), not at all. There's no point in looking that way. It won't even show what boxplots showed in the article
but the relative difference is significant? between a pair of watches with a mean correlation of 1 and another pair with 0.
I don't get it anymore. Let's drop it.
The permutation approach doesn't work (non-intersectional sampling), not at all. There's no point in looking that way. It won't even show what boxplots showed in the article
Somebody did dig it up, though.
Somebody did dig it up, though.
A random coincidence of two curves?
You take one time interval from a random date and the second interval from a random date? What if you shift the starting points?
a coincidence of two curves?
You take one time interval from a random date and the second interval from a random date? What if you shift the starting points?
I've explained everything in the blog.
The blog has it all laid out.
I haven't looked into the code, but if by corr. on non-overlapping samples, then the same fitting, depending on the initial reference points, in addition. Based on trivial logic.
You can get as many such fits as you want when bruteforcing. It's certainly better than nothing, but it doesn't rely on any seasonality whatsoever
Take the 1st and 2nd intervals and randomly mix each of them, see the corr. Someday you will come to an optimum through this approach, but it will have nothing to do with seasonal regularity.
imho