Multiple independent sample testing with proportions

[jamesv87]

Hey everyone,

I'm working on my undergraduate thesis in biology and am having problems analyzing the data from my experiments. I'm wondering if anyone knows how to compare multiple independent groups when the measurements are proportions. The problem is that many of my values are 1 and 0 with relatively few in between. I've tried an arcsine transformation but it does not really help the distribution (Does it matter whether I use degrees or radians?). I've used Kolmogorov-Smirnov tests to compare between two groups at a time, and do find significance, which is reassuring, but I don't know if this is the best way to go about analyzing all of my data. Would I also be violating some rule of science if I did this, similar to using a repeated series of t-tests?

Your feedback is much appreciated!

 

[Enuma_Elish]

You can use probit or logit, with the left hand variable being the proportions pooled across groups, and the right hand variables a set of "dummy" variables that indicate each respective group.