Skip to content

Help with mixed-effects model in lme

3 messages · Jonathan Myers, ONKELINX, Thierry, Dunbar, Michael

#
Dear Jonathan,

I would move the block factor from the ramdom effects to the fixed
effects. You have only two levels of that, which can give rather
unprecise estimates of the variance of the random effects. Moving block
to the fixed effects will cost you only one degree of freedom. So that
would not be a big problem.

lme(species.richness ~ water*fuel*seed*year + block, random = ~1 |plot) 

HTH,

Thierry


------------------------------------------------------------------------
----
ir. Thierry Onkelinx
Instituut voor natuur- en bosonderzoek / Research Institute for Nature
and Forest
Cel biometrie, methodologie en kwaliteitszorg / Section biometrics,
methodology and quality assurance
Gaverstraat 4
9500 Geraardsbergen
Belgium
tel. + 32 54/436 185
Thierry.Onkelinx at inbo.be
www.inbo.be

To call in the statistician after the experiment is done may be no more
than asking him to perform a post-mortem examination: he may be able to
say what the experiment died of.
~ Sir Ronald Aylmer Fisher

The plural of anecdote is not data.
~ Roger Brinner

The combination of some data and an aching desire for an answer does not
ensure that a reasonable answer can be extracted from a given body of
data.
~ John Tukey

-----Oorspronkelijk bericht-----
Van: r-sig-ecology-bounces at r-project.org
[mailto:r-sig-ecology-bounces at r-project.org] Namens Jonathan Myers
Verzonden: donderdag 17 september 2009 22:52
Aan: r-sig-ecology at r-project.org
Onderwerp: [R-sig-eco] Help with mixed-effects model in lme

Dear List Members,

I am using a mixed-effects model in lme and would like to know whether I
am using the proper structure for the random-effects component of the
model. My experiment consists of three categorical treatments (fire,
water, seed) arranged in a split-plot design. The fire treatment (2
levels) and water treatment (3 levels) were assigned to plots, and the
seed treatment (2
levels) was assigned to two subplots within each plot. There are 60
total plots (120 total subplots), divided equally among two large blocks
(30 plots per block). I measured species richness in each subplot in
three separate years. My goal is to test for main effects of the three
treatments, a main effect of year, and all interactions. My current
model consists of four factorial fixed effects (fire, water, seed, year)
and 1 random effect (block), with plots nested within blocks (to account
for the split-plot structure of the experiment):

model = lme(species.richness ~ water*fuel*seed*year, random = ~1 |
block/plot)

The ANOVA output includes two denominator degrees of freedom (denDF): 53
denDF for plot factors (fire, water, fire x water interaction) and 270
denDF for split-plot factors (everything else).

I would greatly appreciate feedback as to whether the random-effects
component of the model looks appropriate, and if not, how it should be
modified.

Thanks very much!

Cheers,

Jonathan


~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
Jonathan A. Myers
Department of Biological Sciences
Division of Systematics, Ecology, and Evolution Louisiana State
University Baton Rouge, LA 70803 USA

E-mail: jmyer19 at lsu.edu
Telephone: 225-578-7567
Fax: 225-578-2597

Website: http://www.biology.lsu.edu/labpages/harmslab/jmyers/index.html
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~


_______________________________________________
R-sig-ecology mailing list
R-sig-ecology at r-project.org
https://stat.ethz.ch/mailman/listinfo/r-sig-ecology

Druk dit bericht a.u.b. niet onnodig af.
Please do not print this message unnecessarily.

Dit bericht en eventuele bijlagen geven enkel de visie van de schrijver weer 
en binden het INBO onder geen enkel beding, zolang dit bericht niet bevestigd is
door een geldig ondertekend document. The views expressed in  this message 
and any annex are purely those of the writer and may not be regarded as stating 
an official position of INBO, as long as the message is not confirmed by a duly 
signed document.
#
Dear Jonathan

A couple of things.

If your have treatments applied to sub-plots then it makes sense to have subplot in the random component, i.e. random = ~1|plot/subplot.

The other issue you have is ALL interactions. With four factors, this is an awful lot of interactions which together will eat a lot of degrees of freedom which would be better in your residual. In the most extreme case, unless you have replicate samples within each year and sub-plot, you may not even be able to estimate the four way interaction correctly. Are you sure that if you see a one of the three way interactions or the four way interaction that you can explain what it means ecologically? 

cheers
Mike


-----Original Message-----
From: r-sig-ecology-bounces at r-project.org [mailto:r-sig-ecology-bounces at r-project.org] On Behalf Of Jonathan Myers
Sent: 17 September 2009 21:52
To: r-sig-ecology at r-project.org
Subject: [R-sig-eco] Help with mixed-effects model in lme

Dear List Members,

I am using a mixed-effects model in lme and would like to know whether I am
using the proper structure for the random-effects component of the model. My
experiment consists of three categorical treatments (fire, water, seed)
arranged in a split-plot design. The fire treatment (2 levels) and water
treatment (3 levels) were assigned to plots, and the seed treatment (2
levels) was assigned to two subplots within each plot. There are 60 total
plots (120 total subplots), divided equally among two large blocks (30 plots
per block). I measured species richness in each subplot in three separate
years. My goal is to test for main effects of the three treatments, a main
effect of year, and all interactions. My current model consists of four
factorial fixed effects (fire, water, seed, year) and 1 random effect
(block), with plots nested within blocks (to account for the split-plot
structure of the experiment):

model = lme(species.richness ~ water*fuel*seed*year, random = ~1 |
block/plot)

The ANOVA output includes two denominator degrees of freedom (denDF): 53
denDF for plot factors (fire, water, fire x water interaction) and 270 denDF
for split-plot factors (everything else).

I would greatly appreciate feedback as to whether the random-effects
component of the model looks appropriate, and if not, how it should be
modified.

Thanks very much!

Cheers,

Jonathan


~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
Jonathan A. Myers
Department of Biological Sciences
Division of Systematics, Ecology, and Evolution
Louisiana State University
Baton Rouge, LA 70803 USA

E-mail: jmyer19 at lsu.edu
Telephone: 225-578-7567
Fax: 225-578-2597

Website: http://www.biology.lsu.edu/labpages/harmslab/jmyers/index.html
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~


_______________________________________________
R-sig-ecology mailing list
R-sig-ecology at r-project.org
https://stat.ethz.ch/mailman/listinfo/r-sig-ecology