How to analyse proportion data that are larger than 1

I've conducted a field experiment testing the effect of herbicide application timing on the level of weed infestation (number of weed per m^2), with control. Now I'm trying to analyse the proportion from control of each treatment over time. However, for some repeats the initial infestation was higher that that of the relevant control, and I get proportion value higher that 1. naturally, arcsine transformation is no use here. Have you any idea how to approach such data?


Active Member
proportion from control = weeds per area treated / weeds per area control?

if it be so:
1) Probably should use poisson regression or related for comparison/modelling
2) or, log xform weeds per area and ANOVAs away.

2) is probably better from practical perspectiv.