Commodity Correlation Table

[late apex]

Are you suggesting that the high positive correlation of, say, corn and Dow (as mentioned above) over the last 180 days is anything but a random occurrence?

Or did you mean something else when you said:

Quote from KS96:

There are no such a thing as a "random correlation".
There is either correlation or not.

The basic point of my reply wasn't about "the books" or any statistical significance testing. This is about a large correlation matrix (1,000+ elements) across a variety (here 35) of diverse asset classes and instruments. Given those two facts, you cannot not have multiple instances of a correlation near one of the two extremes, 1 or -1. And?... And many of those instances will be random (contrary to your statement); an accidental fluke; of no particular meaning, relevance or use, in trading, investing or any other area.

Quote from KS96:

Everything allowed in the books, but my experience says that you shouldn't ask whether a correlation exists by chance or not.

If you are going to make any use of that correlation -- to draw any valid inferences from it -- then that's exactly the question you should ask. Otherwise, it's garbage in, garbage out.

 

[maxforce]

Perhaps a longer period would verify if this is a random correlation or not?

 

[kevinmr]

Quote from KS96:

There are no such a thing as a "random correlation".
There is either correlation or not.

I think what the poster meant is if you looked at a large number of 180 day periods the correlation calculated for each sample would not have been as high as the current one; thus he used the word "random" maybe spurious is a better word.

 

[kevinmr]

Quote from FullyArticulate:

You people don't know what you're talking about.

Well-paid people can now afford to substitute high-grade Corn for the cheaper feed they usually eat.

Isn't corn as cheap as it gets?

 

[FullyArticulate]

Quote from kevinmr:

Isn't corn as cheap as it gets?

I understand the KFC in Greenwich Village is offering some very attractive discounts.

 

[jack hershey]

Check out John Carter and especially pick up his last presentation at Trader's Expo for C, W and S.

 

[osorico]

In my experience, when using limited data, it is very useful to use a "null hypotheses" analysis.

In this case, the null hypotheses is the data represented in the graph itself. One would need several more data sets of past 180 day periods, or at very least, a break down of the existing data into 30 day periods in order to statistically quantify.

I predict, even with a breakdown of existing data, the data will minimalize the high correlation of DOW/Corn. Using additional historical data, the high correlation will be refuted, imo.

Therefore, it is a marginal probability... the high correlation is statistically insignificant because the correlation does not need to exist for the probability of such movement to exist.

My teacher would be proud.