|
Dear all
I wrote about this model
earlier but the problem has not been solved - so here I try again with what I
have so far.
My experiment was carried out
to test whether there are any differences in the functional characteristics
between different cultures within a species of fungus and how big these
differences might be in comparison to differences between species.
However, as very often
with these types of experiments, the design is not balanced for reason of
availability of fungal cultures. Thus Species 1: 13 cultures, Species 2: 5
cultures, Species 3: 4 cultures & Species 4: 2 cultures. Each culture
had 3 replicates i.e. Species 1: 13*3=39 observations in
total.
What we have been trying to do
is to use a nested model with cultures nested within species and after som
discussion we agreed to use cultures as a random factor in a linear mixed
effects model.
We would like to
test/know:
1) whether there are any
differences between species
2) if possible whether there
are any differences between cultures within a given species
3) the amount of variation
between species in comparison to within the species (i.e. between
cultures)
Thus we have tried to group the
data and then use the lme function but the output is not as we
expected. Here follows the grouping, lme call and the output I hope someone
can help to set things straight ;)
DW<-groupedData(ShootDW~1|Species/culture, data="">
mixed<-lme(fixed=ShootDW~Species-1,data=""
class=235020509-25032004>culture~1,subset=1:72,
na.action="">
OUTPUT:
> lme(fixed = ShootDW ~ Species - 1, data = "" random
= culture ~ 1, subset = 1:72,na.action =
"">
Linear
mixed-effects model fit by REML
Data: DW
Subset: 1:72
Log-restricted-likelihood: -2.490799
Fixed: ShootDW ~
Species - 1
SpeciesG.
claroideum SpeciesG. mosseae SpeciesG. geosporum SpeciesG. caledonium
1.131
1.40528
1.6435
0.9958333
Random
effects:
Formula: culture ~ 1 | Species
(Intercept)
StdDev:
0.0986836
Formula: culture ~ 1 | culture %in%
Species
(Intercept)
Residual
StdDev: 0.3334697
0.1547823
Number of
Observations: 70
Number of Groups:
Species culture %in% Species
4
24
For the 3rd question we tried
using the Varcomp function but are unsure whether it is the right figures it
returns for the comparison of the amount of variation between species
and between cultures. The following call to varcomp was used. But since
this and the above does not fit together we are unsure of what is right and what
is not.
is.random(expt7183B)<-T
shoot.vc<-varcomp(ShootDW~Species + culture %in% Species, data="" =
1:72, method="reml",na.action="">
OUTPUT
*** Linear Model ***
Call:
varcomp(formula = ShootDW ~ Species + culture %in% Species, data = "" method =
"reml", subset = 1:72, na.action = "">
Variance Estimates:
Variance
Species 0.03481102
culture %in% Species
0.11339044
Residuals 0.02395417
Method: reml
Coefficients:
(Intercept)
1.276291
I hope someone can give me some
advice on whether this is the way to go and what is wrong with my grouping of
data or lme function. Are there any alternative ways to do it we tried Anova but
this did not seem to be able to do the trick.
All the
best
Lisa
Munkvold
Lisa Munkvold
Lisa Munkvold
RISØ National laboratory
Plant Research Department, build. 313
P.O.Box 49 - DK-4000 Roskilde
Denmark
Phone: +45 46774210
<mailto:lisa.munkvold@risoe.dk>
<http://www.risoe.dk>
|