Most non-parametric tests, are performed on ranked data, so you would have to convert the measurement observations to their ranks in the overakk data set: the smallest value gets a rank of 1, the next smallest gets a rank of 2 and so on. How else can I describe about this ranking?
.
i am not sure to understand your question ...
what you say is the correct description of what the rank statistics is ; you may have to cope with some difficulties when coding for ex aequos (there are multiple choices, see https://en.wikipedia.org/wiki/Ranking )
in practice, this is not a problem since most environments for statistical test start from the natural values and perform the relevant recoding for you
.
if you want to do it by yourself and want to know how to treat ex aequos in a particular statistical test, then you should check the implementations guidelines
for instance, here (Wilcoxon signed rank test)
https://en.wikipedia.org/wiki/Wilcoxon_signed-rank_test
the treatment of ex-aequos is specified by
"Ties receive a rank equal to the average of the ranks they span."
.
Dear Nan, generally speaking, nonparametric tests use ranking because they do not assume the data is scale.
Note, too, that the Wilcoxon Mann-Whitney is not a nonparametric test. It estimates a very useful parameter: the probability that an observation in one sample will be larger than an observation in the other sample. For example, this translates into the probability that a man will earn more than a woman, or that a person on the new treatment will do better than a person in the control group. So although you have lost your original values in the ranking process, you have picked up a very useful measure of effect size!
- Badges
- Science topic