Dear R users:
I've a problem with lme function, when I want
to model an unbalanced two-way anova, with 2 random factors say t and b.
My two models are:
model1- y(ijk) = beta+b(i)+t(j)+epsilon(ijk)
model2- y(ijk)= beta+b(i)+t(j)+b:t(ij)+epsilon(ijk)
beta overall mean effect
The data.frame is X
t b med celda
1 1 10 1
1 1 12 1
1 1 11 1
1 2 13 2
1 2 15 2
1 3 21 3
1 3 19 3
2 1 16 4
2 1 18 4
2 2 13 5
2 2 19 5
2 2 14 5
2 3 11 6
2 3 13 6
I try with lme to obtain the variance estimates like with varcomp, for
model1
model-2, sum and interaction effects.
varcomp gives me:
variance estimates:
t 0
b 0
t:b 7.407
residuals 3.8429
I try with lme:
x <- lme(med~ 1, data = X;random = ~ 1 | t+b
or random = ~t+b | celda
random = ~t*b | celda
random = ~ 1 | t*b , method = "ML")
I get "bad groupping" or "singular convergence".
Please, Can anyone tell me how to model, model-1 and model-2 in lme.
Do you know any library for S-PLUS, R in order to get like in SPSS expected
mean
squares if no library, how obtain in S-PLUS, R?
My e-mail is darkjacknife at hotmail.com
_________________________________________________________________
Protege tu correo del spam y los virus con MSN 8. Prueba gratis dos meses
MSN 8. http://join.msn.com/?pgmarket=es-es&XAPID=199&DI=1055
