Hi all,
This is a statistics question, I hope someone out there will be able to help
me.
I have one population (oligonucleotide probes spotted on a Nimblegen array).
I measured "parameter one" (intensity after hybridisation) and I have
selected a subpopulation of the initial (one tenth of the initial) according
to a threshold value.
I now measure "parameter two" of the original population (GC content).
The
distribution of the population for this value is roughly bell-shaped.
I want to see if the subpopulation I selected in the previous step shares
the same characteristics with regards to the GC content with the entire
population or if selecting for "parameter one" has messed with
"parameter
two".
What I thought was to compare the distributions of this second attribute of
the two populations.
I believe that the ansari-bradley, wilcoxon and Kolmogorov-Smirnov tests
perform such tests but -after searching- I am not sure which is more
appropriate (if any).
I realize that ansari-bradley and ks are more sensitive to the actual shape
of the curve while wilcoxon focuses on testing for a shift of the median. I
can not figure out though what is the difference between ansari-bradley and
ks . Is there any important difference in the assumptions of these three
tests that I should consider before choosing?
Finally, and I apologize for the naivity of the question, all the ks.test(),
wilcox.test () and ansari.test() expect the raw measurements for the
populations and I do not need to pre-process in any way, right?
ANY suggestion please?
Niki
[[alternative HTML version deleted]]
