How to test whether percent values differ according to a time condition? I'm currently stuck on this data, trying to figure out what test is needed. I started thinking it needed an ANOVA test but on reading, figured it was a Chi Squared test. On more reading, that doesn't seem right either.
The data I have is in a table as such:
Damage Group    0 minutes     2 minutes
11              40.17         28.38
12              10.08         9.47
13              9.30          12.12

and so on. The data in the minute columns represent percentages of the results over the damage group. (Damage Group has 9 levels, I haven't shown them all and there's 15 columns of minute data)
I want to see if the different time distributions are significantly different to each other. What test is needed for this?
 A: Yeah, you should do an ANOVA to determine if any of the groups are different from each other.
Then, if they are, you need to start looking at what ones are likely to have a significant difference.
You need to be careful testing so many groups as it is easy to get false positives, Tukey's HSD test is good for this (also the multicomp function in S+). Are you using any particular software for the analysis? Let me know and I'll tell you if I can help with any code.
http://en.wikipedia.org/wiki/Tukey%27s_range_test
A: You should start by peforming an ANOVA if you are trying to determine if there are differences in any of the mean responss for each group.  You'll need to check the assumptions of the ANOVA model by creating graphs and plots and other diagnostic tests to determine the feasibilty of the assumptions (such as normality of the error terms).  If you find the data violate the assumptions, then you will need to either transform the data to satisfy the assumptions or (or use other remediation methods) or you can perform a nonparametric ANOVA Kruskal-Wallis test.
