Dears useRs,
I have 2 factors, (for the sake of explanation - A and B), with 4 levels each.
I've already fitted a negative binomial generalized linear model to my data,
and now I need to split the factors in two distinct analysis of deviance table:
- A within B1, A within B2, A within B3 and A within B4
- B within A1, B within A2, B within A3 and B within A4
Here is a code that illustrates my problem:
# inputing my data
require(MASS)
my.data <-
data.frame("A"=rep(c(rep("Alevel1",4),rep("Alevel2",4),rep("Alevel3",4),rep("Alevel4",4)),4),
"B"=c(rep("Blevel1",16),rep("Blevel2",16),rep("Blevel3",16),rep("Blevel4",16)),
"value"=rnegbin(64, 10, 10))
# fitting the model with interaction (a + b + a:b)
model <- glm.nb(value ~ A*B, data=my.data)
anova(model, test="F")
Df Deviance Resid. Df Resid. Dev F Pr(>F)
NULL 63 80.639
A 3 1.374 60 79.265 0.4581 0.7115
B 3 2.285 57 76.980 0.7616 0.5155
A:B 9 9.700 48 67.280 1.0778 0.3753
#until here it's ok, now I need to redistribute the 12 degrees of freedom
(B+A:B -> 3+9) into 4 splitted factors (A within B1, A within B2...)
model2 <- glm.nb(value ~
A/((B=="Blevel1")+(B=="Blevel2")+(B=="Blevel3")+(B=="Blevel4")), data=my.data)
anova(model2, test="F")
Df Deviance Resid. Df Resid. Dev F Pr(>F)
NULL 63 80.639
A 3 1.374 60 79.265 0.4581 0.7115
A:B == "Blevel1" 4 3.871 56 75.394 0.9676 0.4238
A:B == "Blevel2" 4 5.236 52 70.158 1.3090 0.2639
A:B == "Blevel3" 4 2.878 48 67.280 0.7195 0.5784
A:B == "Blevel4" 0 0.000 48 67.280
However, the 12 degrees of freedom are distributed as 4,4,4,0 and not 3,3,3,3,
as I expected.
Is there a way of obtaining the split I need?
Thanks in advance, all the best!
Rafael.
____________________________________________________________________________________
[[elided Yahoo spam]]
[[alternative HTML version deleted]]
______________________________________________
R-help@r-project.org mailing list
https://stat.ethz.ch/mailman/listinfo/r-help
PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
and provide commented, minimal, self-contained, reproducible code.
