problems with Analysis Of Variance of unbalanced non parametric data

Hello fellow R users :wave:

I hope someone can help me solve this problem. I've been looking for a solution all day.

I have a proportional response (% cover) variable.
I have 2 explanatory factors, both with 3 or more levels: quadrats (of which there are many) and year (of which there are 3).

Having eventually ruled out a Friedmans and Skillings-Mack tests I came back to GLMs.

I was trying
model<-glm(cover~quadrat*year, family=binomial)

However R keeps returning warnings. I tried to balance my data by adding many "NA"s where data wasn't collected but one error is still coming up:
Warning message:
In eval(expr, envir, enclos) : non-integer #successes in a binomial glm!

If anyone can shine some light on where I am going wrong that would be greatly appreciated!

Many Thanks in advance :)



Ambassador to the humans
The response needs to be the actual success counts and the number of trials (if it's not just 0/1 response). There is a big difference between 1 out of the 2 trials being successful and 1000000000 out of the 2000000000 trials being successful and glm needs to know that information.
Thanks for replying Dason,

The response isn't a success-failure type of variable. It is a percentage cover of an area - I have it in decimal form - 70% or 0.7 so I'm not sure if your example relates?


Ambassador to the humans
If that is the case then using logistic regression (binomial family) isn't possible. Something like a Beta regression might be more appropriate.