Conditions: Before drug; after drug (within-cell comparison)
Parameters measured: X, Y
N=5 [and a hard-won 5, at that]
X increases in all cells
Y increases in all cells
Parameter X, Wilcoxon paired test: 0.039
Huzzah! Glory, publication, and scientific knowledge are mine!
Parameter Y, Wilcoxon paired test: 0.06
Boo. Boo. Many times boo.
Seriously though, how is this possible? Now I have to get another n, despite the fact that it is obvious to any clod that Y increases following drug application. Corollary question: how is it possible that I love math, but hate stats? (especially those which seem designed to thwart the earnest scientist....i.e. all of them).
15 years ago
4 comments:
Erm, ew? I know basically nothing about stats...i seem to recall a t test or something like that... Good luck!
You might enjoy looking at some of my posts in the Chemiotics subsection of "The Skeptical Chymist", particularly the first one. Comments are always welcome.
Retread
You know the old adage, there are lies, damn lies and statistics. In your case, the stats are just lying to thwart your work. I'm just glad you are honest enough to go after that extra boost in n, some less scrupulous people would just make it up (I've been bitten by others doing this before). At least the experiment didn't work the last time, completely ruining your stats!
I'd argue p=0.06 is plenty significant. biology isn't an exact science, it's complicated and noisy and can't start with assumptions like a spherical bird or a frictionless surface with no gravity. p <0.06 is just a little bit worse than p<0.05. I have had success reporting p=0.06 results -- they are significant at the 6% alpha, and that's not that far from the 5% alpha. but that work would have been much harder to get more replicates of, and impossible to block them experimentally to match the others. So perhaps this acceptance is restricted to ecology and evolutionary biology.
Post a Comment