RCBD Analysis

#1
I have a randomized complete block design (RCBD). I have 2 treatments (A and B). I have 3 blocks (3 watersheds) of each treatment (A1, A2, A3 and B1, B2, B3). Within each block I have 15 plots (10/10meters). Within the 15 plots I collected data each time I found one of 3 different types of bushes (M, E, and R) in my plots. I need to run an analysis that will tell me if treatments A and B have a significantly different number of Bushes (1, 2, and/or 3) or if the amounts of each bush type have a significant relationship with treatment. My data looks like a row for each individual bush with a column for treatment (B or U), a column for block(A1, A2, A3, B1, B2, or B3), a column for plot # (1-15), and a column for the bush type (M, E, or R) Any help would be really appreciated! I am a novice at stats so the more detail the better. My main problem is I don't know what needs to be (if any) nested or crossed.
A sample of my data is below, but I have it for 398 bushes:
 

katxt

Well-Known Member
#2
In block B1 Plot 1 (top line example) you found 3 M type bushes, there would be 2 more rows exactly like this row?
Or is it presence/absence data and those 3 bushes are included in row 1?
 

katxt

Well-Known Member
#3
This is how I read things.
There are 3x2x15=90 plots. For each plot you record 3 numbers (the count of M, R and E).
So your full data is 270 rows deep and 5 columns wide including 4 predictors and the response.
Block, treatment and bushtype are all crossed because you have data for every combination.
Plot is nested in treatment because a single plot does not have all three treatments.
Plot is nested in block for a similar reason.
The total of all the responses from the 270 rows is 398 so each response going to be mainly between 0 and 3 say.
Your responses mean that a normal anova won't work. You will probably need to investigate a general linear model with Poisson responses.