[R-sig-ME] specifing "error b" to a split plot design in a GLS model

[Javier Moreira]

hi,
i have a split plot design with the bigger plot set as "AMBIENTE", i had
been working with mlm but, in order to campare, im trying a model with no
random effects.
So, in that case, at least en nlme package have to be fit with GLS.

i copy the model,


but, what i need to do is that the sum of squares of AMBIENTE had to be
compare not with the general error, but with the one represented by the
BLOQUE*AMBIENTE interaction.
Can anybody help?

thanks so much for your time.

library(nlme)

attach(BASE_MODELO1_1_1)
str(BASE_MODELO1_1_1)
data2=BASE_MODELO1_1_1

names(data2)=c("ID","BLOQUE","TRATAMIENTO","AMBIENTE","RTO","Ntot","POBLACION","SPAD")

modelo1<-gls(RTO~1+TRATAMIENTO*AMBIENTE+BLOQUE:AMBIENTE,
                  data=data2,
                  method="ML")
summary(modelo1)
anova(modelo1)

​
 BASE_MODELO1_1.1.csv
<https://drive.google.com/file/d/0B6YImu-ZATe4bkRRblRpODFQMXc/view?usp=drive_web>
​
-- 
Javier Moreira de Souza
Ingeniero Agrónomo
099 406 006

	[[alternative HTML version deleted]]