In my experiments I tried different set ups to balance the distribution between two tasks. Each set up was run 32 times. I got the following task distributions [ratio from 0 to 1 of tasktype1/(tasktype1+tasktype2)]: http://oi63.tinypic.com/2cf6szb.jpg
This is how (part of) the dataframe looks like in R: http://oi67.tinypic.com/2z9fg28.jpg
I think ANOVA is not suitable as there seems to be no normal distribution of the data. (Is there a quick way to verify a low level of normality? Is there a standard at which point ANOVA is not suitable anymore?)
Therefore I decided to do the Kruskal-Wallis test. Reading on the test I figured the data needs to be ranked. But how can I choose the method of ranking when computing the Kruskal-Wallis test in R. In my case the "desired" outcome is a balanced population (ratio of 0.5). So the ranks would be:
rank: ratio:
1 0.5
2 0.4 and 0.6
[...] [...]
Can kruskal.test() be adjusted accordingly? (Maybe I am understanding the function wrong...)
My best guess would be just to try: kruskal.test(ratio ~ Method, data = ds)
