Educate me on the correl function

=CORREL(y values, x values)
y values would be test scores, x values would be time or log time or days between tests.
CORREL gives the correlation between linear curve (from your x and y values) versus the actual data. A 1.00000 means perfect correlation. The higher (maximum of 1) the better. The Correlation Coefficient is usually designated "R" in math.
Does that help?

Are you concerned mostly with the Excel implementation, or are you looking for some broader discussion about the Pearson correlation coefficient (the quantity computed by the Excel CORREL() function).

If you need help understanding the concept statistically, I would probably suggest http://www.real-statistics.com/correlation/ and/or http://en.wikipedia.org/wiki/Pearson...on_coefficient and/or http://stattrek.com/statistics/correlation.aspx and/or put "pearson correlation coefficient" or similar into your favorite internet search engine.

If you already understand how the Pearson correlation coefficient is calculated and used, and just need some discussion around Excel's function (examples of use, descriptions of arguments, etc.), I would probably start with the help file for the CORREL() and/or PEARSON() functions found in your installation's help files.

It still seems like we are asking statistics questions and not really Excel questions. We would be having the same discussion points whether we were doing this in Excel, or Minitab, or Matlab, or on a handheld calculator, or using a slide rule.

I did a correlation between Minutes Studied to Test Scores and got a value of -0.275790235. I understand that correlation does not mean causation, but based of this correlation we can assume more time spent studying is actually a bad thing, as the more time spent studying will result in worse test scores.

You are correct in seeing that there is a negative correlation here. As you say, the interpretation of why there is a negative correlation could easily be off. Perhaps it shows, instead, that those who understood the material did not waste as much time studying ("I already understand this, so I'm not going to keep studying"), while those who had a harder time grasping the material spent more time trying to understand the material.

I also did a correlation between average minutes per study session to test scores and got -0.715034083. It has been a while since I was in a statistics class but this seems like a significant value.

In an "off the top of my head" kind of way, yes, a correlation coefficient of -.7 seems significant. Much of the science of statistics is about quantifying this idea. I am not off the top of my head familiar with these tests, but I'm almost certain that there are "significance tests" around the correlation coefficient. The Wikipedia article has a brief section about "testing using student's t distribution", suggesting to me that a t-test could be structured to determine (within your desired confidence interval) whether or not a given coefficient was truly not equal to 0. If you are looking for this kind of hypothesis testing using the correlation coefficient, I would suggest further research into how these tests are set up and performed.

Since Correl ignores text, it seems to me a number would have to be assigned to each location, so library would be 1, home would be 2 and coffee shop would be 3. Doing it this way, would an accurate correlation between study location and test scores be figured?

I'm afraid I am not enough of a statistician to say how one would interpret a correlation coefficient based on these kind of "category" variables. Off the top of my head, I would be tempted to assign a "noise level" or "distraction level" score to each location in the study (or some other meaningful quantity that I think might help explain how location figures into the results), and then test my correlations against that score. I'm sure that there are other ways to deal with this kind of "category" variable, I just am not familiar enough with the statistical science behind this kind of correlation effort.

Correlation is specifically relating a dependent variable y to changes in an independent variable x. The measurement of x must be on a specific scale, i.e. all in minutes of study time (can't have one in seconds, next one in days, etc.). To that point, Mr. Shorty's suggestion of attaching a "level of distraction" to each location would work while arbitrarily attaching a value of 1,2,3 would not.