My problem is this: I have a need to do some analysis of student test results, and I am seeking some tools in order to do so. What I have are collections of scores for a number of different tests, and a number of different students. So each student may have zero or more scores, and each test may have zero or more results. For some test (let's say Test A) I have 30 scores, distributed across the possible range (1 to 9). For a second collection point I have a similar result set; the scores are distributed differently, but still across the same range.
So, I know I can calculate the mean and look at those. But that doesn't tell me a huge amount. Essentially I want to find (1) some way to compare one collection to another, and (2) a way to quantify it to a numerical value; in this way I could say "the amount improved between collections 1 and 2 was twice as much as that between collections 2 and 3" - or something similar. I've read a bit about effect size, but I'm not sure if it will suit my purpose.
Additionally, each test conforms to a different range and scale; I have found an algorithm that works for me to 'normalize' the scores into a common base. Provided the algorithm is perfectly weighted (for argument's sake), would it be safe to use data from multiple different tests all together as one 'collection' of scores?
To sum it all up, I suppose I am looking for any tools that would be useful in my endeavor to quantify and compare data sets of test scores.