Three Variable Correlation

[Eileen]

Hi.

I was wondering if it was possible to work out the correlation between three
sets of variables in excel, and not just two.

The data looks like this (example values):

15,283,741,924 22,891,175,671 182,489,383
12,450,453,827 25,604,121,237 458,039,634
25,017,912,202 10,105,478,615 408,312,030
15,608,597,275 27,352,700,295 71,393,406
6,293,126,663 33,730,623,537 699,182,754
15,665,042,562 25,952,123,256 76,238,591
11,565,224,759 30,426,279,924 44,724,721
8,716,353,737 33,395,986,265 135,495,541
11,234,337,068 23,463,987,625 1,163,192,672
16,285,868,769 21,238,873,761 1,777,205,148
10,406,955,459 31,623,688,840 71,934,583
5,827,079,717 38,907,282,319 964,824,980
11,626,458,649 21,043,475,527 35,381,302
4,081,044,992 41,239,034,015 835,902,380

I've used the correl function to work out the two variable correlation
between columns A/B, A/C and B/C. Now I'd like to see the three variable
correlation between columns A/B/C.

Is this possible?

Thanks in advance for your help.

 

[Ben Evans]

Hi,

Mathematically no such measure exists. Correlations are a measure (if you will) of relation between two variables ONLY. Measuring the three correlations separately is the typical approach. You could run PCA (principle component analysis) on your correlation matrix to understand their linear relationship somewhat more.

If you are in interested in understanding and modelling the relationship between more than two variables OR are interested in other modelling techniques that measure the relationship between two variables then I suggest that you research the topic known as "Copulas". This will also introduce you to the next step of random variable relationships and their lovely complexities.

Mathematically Copulas can be a substantial step so only tread there if mathematically able and willing to learn.

Cheers,