Jonathan Coop's advice -- linear model with Poisson data-- is good. I would add 
that I think you are 
approaching your data the wrong way -- you admit that your zeros are good data, 
and so they are 
neither a 'plague,' nor something that can, or should, be 'cleaned up.' Find 
the statistical model that 
fits the data, rather than the other way round.

Gareth Russell
NJIT/Rutgers

Reply via email to