Finding correlation between two percentages I have lots of graphs that look similar to this one:

I want to find the 5 graphs where the correlation between the green and blue bars is the greatest.
I want to measure the correlation between the green and the blue bars on this graph, if there is one. The green bars sum to 100% and the blue bars also sum to 100%.
I had thought of measuring the difference between each pair of blue and green bars, summing it and squaring it. I don't remember much from my statistics class but I think this is called linear regression (please correct me if I'm wrong).
I was going to do this for all of my graphs and pick the 5 where R squared is the smallest. Is this correct and if not, what method should I use instead?
Thanks,
Francis
EDIT:
Thanks to everyone who's helped so far. I think my original question was worded incorrectly. When I said I wanted to find the correlation between the green and blue bars, I meant I want to find the correlation between the counters and time variables.
 A: Sample correlation is measured for a dataset containing a set of pairs of variables, and it requires at least two pairs of variables to be well-defined (each pair is essentially a single datapoint, and we need at least two of these).  Thus, it is possible to compute the sample correlation between counters and time for the entire dataset you have shown, but it is not possible to compute correlation for a single pair of values.  You could pick the point where both values make the greatest contribution to the sample correlation for the overall data-set, but I do not see what that accomplishes.
I'm afraid that this appears to be one of the "false premise" questions we get a bit on this site, where the questioner is asking of advice on how to implement a procedure that either does not exist, or does not constitute a good method of analysis to begin with.  Rather than asking how you might implement some arbitrarily constructed procedure, it would be more useful if you were to ask a new question explaining the overall research/statistical question you have for this data, and seeking advice on how best to deal with this.  Good luck.
