Correlation Analysis

Quote from ImamicPH:

Let's say I wanted to test two commodities to see how much they correlated over time.

What's the starting point?


If one does not have the foggiest about correlations, the starting point is to use a charting program or spreadsheet to calculate and display the correlation.
 
Quote from culion:

If one does not have the foggiest about correlations, the starting point is to use a charting program or spreadsheet to calculate and display the correlation.


It's not that I don't know how to find correlation. I do this for a living.

I mixed up the term correlation as defined by math and science, and the term "correlation" as we throw it around our office.

I think word I was intended was "divergence in terms of percentage of price".

These two used together may be used to screen for trade setups.

I've got the correlation done for about ten different markets.
 
correlation ranges from -1 to +1. perfectly negative correlation (-1) means the two move in complete opposite direction ( X gains 1%; y loses 1%). perfectly positive correlation means they move in complete synch with each other...if X gains 1%, y gains 1%.
 
Another factor you can look at is the correlation of the equity curves for a given system. If you are trying to build a portfolio of lowly correlated markets to test a basket system, then correlation of the equity curves is the best measure. You might find it hard to believe but , for a ORB system the SP500 and Nasdaq correlation is in the .2 to .3 range, not as high as you would think.
 
Quote from rosy:

you correlate based on returns NOT price levels.

Yes, I am using correlation to build a portfolio which has it's members not highly correlated on a equity curve basis.

I discuss this correlation report half way down this page
http://www.tradersstudio.com/Overview/tabid/68/Default.aspx?PageContentID=9

I will upload a sample workbook with this report in it so everyone can see it.
 
Quote from Murray Ruggiero:

Yes, I am using correlation to build a portfolio which has it's members not highly correlated on a equity curve basis.


I'm a little thick tonight -- what does that even mean?
 
I simply do correlation analysis on the daily, weekly or monthly changes in the equity curve of the system.
 
Back
Top