To test fixed effects you can do anova or AIC comparisons or parametric bootstrap to compare two nested models (all using ML rather than REML).
regards
Mike
-----Original Message-----
From: ONKELINX, Thierry [mailto:Thierry.ONKELINX at inbo.be]
Sent: 21 January 2010 09:17
To: Zoltan Botta-Dukat; Dunbar, Michael
Cc: r-sig-ecology at r-project.org
Subject: RE: [R-sig-eco] FW: mixed effect model: compare seed families
Dear Zoltan,
lmer not displaying p-values is FAQ 7.35 (http://cran.r-project.org/doc/FAQ/R-FAQ.html#Why-are-p_002dvalues-not-displayed-when-using-lmer_0028_0029_003f)
HTH,
Thierry
----------------------------------------------------------------------------
ir. Thierry Onkelinx
Instituut voor natuur- en bosonderzoek
team Biometrie & Kwaliteitszorg
Gaverstraat 4
9500 Geraardsbergen
Belgium
Research Institute for Nature and Forest
team Biometrics & 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: Zoltan Botta-Dukat [mailto:bdz at botanika.hu]
Verzonden: donderdag 21 januari 2010 9:30
Aan: Dunbar, Michael
CC: r-sig-ecology at r-project.org; ONKELINX, Thierry
Onderwerp: Re: [R-sig-eco] FW: mixed effect model: compare seed families
Hi,
Thanks to Mike and Thierry for suggestions. The key points were that the random factors are crossed, and it can be analysed by lmer.
I've done the analysis, and the results considerably differ from which I've get without including random factors.
It was strange for me that lmer calculate t-statistic for each parameter estimates, but it does not show neither df nor p-value.
For example (from the help of lmer):
> (fm1 <- lmer(Reaction ~ Days + (Days|Subject), sleepstudy)) Linear mixed model fit by REML
Formula: Reaction ~ Days + (Days | Subject)
Data: sleepstudy
AIC BIC logLik deviance REMLdev
1756 1775 -871.8 1752 1744
Random effects:
Groups Name Variance Std.Dev. Corr
Subject (Intercept) 612.092 24.7405
Days 35.072 5.9221 0.066
Residual 654.941 25.5918
Number of obs: 180, groups: Subject, 18
Fixed effects:
Estimate Std. Error t value
(Intercept) 251.405 6.825 36.84
Days 10.467 1.546 6.77
Correlation of Fixed Effects:
(Intr)
Days -0.138
I could test factors only by comparing two models using anova function that calculates ML-Chisq (I tried the test="F" parameter, but it did not influence the result). I hope it is correct.
Thanks again, and best wishes
Zoltan
Dunbar, Michael írta:
> Hi Zoltan
>
> Possibly plot and source site are crossed. ie (1|plot) + (1|site) in
> lmer (not nlme)
>
> There would then be no problem in having fixed effects corresponding to sites and properties of individuals. As site is random, you can test sand vs clay, but a test of residual differences between sites is tested by comparing models with and without the site random effect. If you just want to test this within sand then just fit the models to the sand site data, but you then many be running short of replication?
>
> You don't mention your response variable, is each individual measured more than once, of not then you can't specify an individual as random as there is no replication within it.
>
> regards
>
> Mike
>
>
> ________________________________________
> From: r-sig-ecology-bounces at r-project.org
> [r-sig-ecology-bounces at r-project.org] On Behalf Of Zoltan Botta-Dukat
> [bdz at botanika.hu]
> Sent: 18 January 2010 19:26
> To: r-sig-ecology at r-project.org
> Subject: [R-sig-eco] mixed effect model: compare seed families
>
> Hi everyone,
>
> I'm struggling with evaluation an experiment for maternal effect and
> plasticity.
>
> We have grown offsprings of 64 individuals (seed families) in a common
> garden experiment. There was 8 replicate plots, in each plot one
> individual from each seed family.
> Seeds come from 8 different sites, and 8 seed families collected in
> each site. Sites belong to two groups according to soil type: sand and clay.
> In each site 4 large and 4 small mother plant were chosen.
>
> We would like to test:
> - difference between soil types
> - difference between mother size categories
> - interaction between soil type and mother size
> - difference between sites within a soil type (In fact we are
> interested in differences between sites in sand only, because the
> common garden was on sand, and one site is situated in its neighbour).
>
> It is clear that soil_type and mother_size are fix factors.
> I suspect that both plot and seed_family should be random factor, but
> I'm not sure what is the correct specification. Maybe
> random=~1|plot/seed_family. It's OK?
> Can I include the comparison between sites into this model? Or would
> be better to make a separate analysis using seed families from sand only?
>
> Thanks for the suggestions
>
> Zoltan
>
> _______________________________________________
> 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.
--
This message (and any attachments) is for the recipient ...{{dropped:6}}