[R-sig-ME] repeated measures on split plot design
thierry.onkelinx at inbo.be
Mon Mar 16 15:17:16 CET 2015
Much depends on how the observations are correlated. nlme handles
correlated residuals within the most detailed level of the random effects.
In your case it will assume some correlation structure among residuals of
the same block:veg combination. Residuals from different block:veg
combinations are assumed to be independent.
Autocorrelation as a higher level (e.g. block or even the main effects) is
not possible in nlme. The INLA package allows for correlated random effects.
You need to provide more information on the kind of temporal correlation
that you want to incorporate and at which level is should operate.
Note that you have not specified veg. block has only very few levels. It
might not be a good idea to use it as a random effect. See "Should I treat
factor xxx as fixed or random?" on http://glmm.wikidot.com/faq
ir. Thierry Onkelinx
Instituut voor natuur- en bosonderzoek / Research Institute for Nature and
team Biometrie & Kwaliteitszorg / team Biometrics & Quality Assurance
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
2015-03-13 16:47 GMT+01:00 Lea <mwiederm op mtu.edu>:
> I am hoping for help with formulating the random structure for a data set
> that fits a log normal distribution. The data was collected from a split
> plot designed experiment (block: 4 replicates; v treatment: 3 levels; d
> treatment: 2 levels). d is nested within v which is nested within block;
> total n = 4x3x2 = 24; my random effects are all categorical
> hence using lme: random= ~1|block/veg, or lmer: (1|block/veg)
> I think I am OK with that but I can not figure out how to include the
> temporal repeated measure. This split plot setup was samples 5 times
> (monthly during the summer) for 3 years-> 15 times and I am too interested
> in the month and year effect. Ergo I do not want to average it out. Also
> from plotting the data there is an increase of the response throughout the
> Most everybody seems to be using the lme4 and not the nlme package these
> days. Is there still someone who could help me to write the temporal
> structure of the random term using the nlme package?
> my fixed factors are: v*d*year*month (I am not interested in the block
> effect) my fixed effects are all categorical
> m1<-lme(response~v*d*year*month, random= ~1|block/veg and what do I do
> the 15month in 3 years????
> m2<-lmer(response~v*d*year*month+(1|block/veg and what do I do with the
> 15month in 3 years????
> Thank you so much!
> p.s.: I hope I don't offend anybody by writing things out for both
> R-sig-mixed-models op r-project.org mailing list
[[alternative HTML version deleted]]
More information about the R-sig-mixed-models