R help - Jun 2005 - Anohter anova.mlm problem

Hi,

yet another anova.mlm problem - it doesn't seem to end.
This time, I have a setup with a few within-subject factors and a
between-subject factor (SGROUP). Consider the most simple case with only one
within-factor (apo):
> mlmfit0 <- lm(data.n ~ 0 + SGROUP)
> mlmfit1 <- lm(data.n ~ 1 + SGROUP)
> anova(mlmfit1,mlmfit0,test="Spherical",M=~hemi,X=~1)
Analysis of Variance Table

Model 1: data.n ~ 1 + SGROUP
Model 2: data.n ~ 0 + SGROUP

Contrasts orthogonal to
~1


Contrasts spanned by
~hemi

Greenhouse-Geisser epsilon: 0.6829
Huynh-Feldt epsilon:        0.7127

  Res.Df Df Gen.var.   F num Df den Df Pr(>F) G-G Pr H-F Pr
1     22     0.54295
2     22  0  0.54295 Inf      0     44
Warning messages:
1: NaNs produced in: pf(q, df1, df2, lower.tail, log.p)
2: NaNs produced in: pf(q, df1, df2, lower.tail, log.p)
3: NaNs produced in: pf(q, df1, df2, lower.tail, log.p)


Now, those G-G and H-F epsilons it gives me are correct (at least G-G, the
H-F seems to be affected by the R vs SAS difference in the calculation). But
whatever happened to everything else?
I've played around with the source in mlm.R, and it looks to be as though
the problem is related to this:
https://stat.ethz.ch/pipermail/r-devel/2005-April/032925.html

Is there a solution to this problem? Am I doing something wrong? The data
looks like this:
> hemi
 [1] 1 1 1 1 1 1 2 2 2 2 2 2 3 3 3 3 3 3
Levels: 1 2 3
> SGROUP
 [1] 2 2 2 2 2 2 1 1 1 1 2 2 2 1 1 1 2 1 1 1 2 1 2 1
Levels: 1 2
> dim(data.n)
[1] 24 18

It works fine with the univariate analysis (aov())...

Thanks again,

Bela

-- 
Weitersagen: GMX DSL-Flatrates mit Tempo-Garantie!
Ab 4,99 Euro/Monat: http://www.gmx.net/de/go/dsl

"Bela Bauer" <bela_b at gmx.net> writes:
> Hi,
> 
> yet another anova.mlm problem - it doesn't seem to end.
> This time, I have a setup with a few within-subject factors and a
> between-subject factor (SGROUP). Consider the most simple case with only
one
> within-factor (apo):
> 
> > mlmfit0 <- lm(data.n ~ 0 + SGROUP)
> > mlmfit1 <- lm(data.n ~ 1 + SGROUP)
> > anova(mlmfit1,mlmfit0,test="Spherical",M=~hemi,X=~1)
> Analysis of Variance Table
> 
> Model 1: data.n ~ 1 + SGROUP
> Model 2: data.n ~ 0 + SGROUP
> 
> Contrasts orthogonal to
> ~1
> 
> 
> Contrasts spanned by
> ~hemi
> 
> Greenhouse-Geisser epsilon: 0.6829
> Huynh-Feldt epsilon:        0.7127
> 
>   Res.Df Df Gen.var.   F num Df den Df Pr(>F) G-G Pr H-F Pr
> 1     22     0.54295
> 2     22  0  0.54295 Inf      0     44
> Warning messages:
> 1: NaNs produced in: pf(q, df1, df2, lower.tail, log.p)
> 2: NaNs produced in: pf(q, df1, df2, lower.tail, log.p)
> 3: NaNs produced in: pf(q, df1, df2, lower.tail, log.p)
> 
> 
> Now, those G-G and H-F epsilons it gives me are correct (at least G-G, the
> H-F seems to be affected by the R vs SAS difference in the calculation).
But
> whatever happened to everything else?
> I've played around with the source in mlm.R, and it looks to be as
though
> the problem is related to this:
> https://stat.ethz.ch/pipermail/r-devel/2005-April/032925.html
> 
> Is there a solution to this problem? Am I doing something wrong? The data
> looks like this:
You're fitting the same model twice, so you get zero DF for the
difference. If you want to compare a model with a difference between
the two groups to one with no difference (in the relevant contrasts),
you need to compare

 mlmfit0 <- lm(data.n ~ SGROUP)
 mlmfit1 <- lm(data.n ~ 1)


-- 
   O__  ---- Peter Dalgaard             ??ster Farimagsgade 5, Entr.B
  c/ /'_ --- Dept. of Biostatistics     PO Box 2099, 1014 Cph. K
 (*) \(*) -- University of Copenhagen   Denmark          Ph: (+45) 35327918
~~~~~~~~~~ - (p.dalgaard at biostat.ku.dk)                  FAX: (+45) 35327907