What is an appropriate technique for measuring correlation between classes? I am a school teacher doing a small project that involves collecting data on 5 different teaching methods and the resulting test score of children who are taught was each teaching method. I have 50 students equally split among the 5 teaching methods (10 students/method). The data I have is as follows. Each line below is one student where the first column of data is the teaching method and the second column is the student's score:
Method A,88
Method B,72
Method A,90
Method D,55
Method D,60
Method A,84
...

I would like to somehow determine if test scores correlate with the different methods. I've been reading about Intraclass Correlation and I want to know if Cohen's kappa or a similar technique would be an appropriate way to look at the correlation of test scores and teaching methods. Any help or direction is much appreciated.
 A: You are essentially interested in whether your teaching methods (a grouping variable) have an impact on test scores. If test scores within methods were normally distributed, with roughly equal variances, you could use one-way analysis of variance. However, I suspect that these conditions are not met, so a Kruskal-Wallis test would be appropriate. Good luck!
A: I'm not a mathematician, nor do I use statistics.
As a scientist, though, any conclusions you can draw from the data will be far more robust if you use all five methods of instruction on all five groups of students during different lessons.
1) This will be more equitable to all the students if it turns out one method is particularly bad or good.
2) You will then have exactly one basic group, with the same general deviation, and can then run the appropriate statistical technique and draw more accurate conclusions as to cause and effect.
Any teaching method that is beneficial to one student is almost guaranteed to be neutral or harmful to another student relative to that student's preferred methods (i.e. individual learning, mentor, group approaches).  Regardless of "learning style".  So please try to factor this in to your analysis.  Schooling is the only form of psychological experimentation in which the experimenters don't need informed consent from the subjects.  This alone carries with it a profound ethical obligation.
