[R-sig-ME] Comparing variance components of crossed effects models fit with lme4 and nlme
jmichaelrosenberg at gmail.com
Mon Aug 14 18:23:23 CEST 2017
Thank you, I will explore use of INLA (or potentially the brms package
because of my familiarity with the [lme4-like] syntax).
I'm curious whether you (or anyone else) has thoughts / advice on using a
package that uses a Bayesian approach for carrying out mixed effects
modeling. In my field / area of research, mixed effects models are new! And
so a Bayesian approach to them would be *very *new. Even though if I
understand (very preliminarily), with some (uniform) prior specification,
results can be comparable to models specified with a maximum likelihood
approach, when possible.
Thank you again!
On Fri, Aug 11, 2017 at 8:38 AM, Thierry Onkelinx <thierry.onkelinx at inbo.be>
> Dear Joshua,
> Crossed random effects are difficult to specify in nlme. I think that you
> have to use pdBlocked() in the specification.
> When I need correlation I often use INLA (r-inla.org). It allows for
> correlated random effects. Crossed random effects are no problem.
> Best regards,
> ir. Thierry Onkelinx
> Instituut voor natuur- en bosonderzoek / Research Institute for Nature and
> team Biometrie & Kwaliteitszorg / team Biometrics & Quality Assurance
> Kliniekstraat 25
> 1070 Anderlecht
> 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
> 2017-08-10 23:05 GMT+02:00 Joshua Rosenberg <jmichaelrosenberg at gmail.com>:
>> Hi all,
>> I'm trying to fit models with a) crossed random effects and b) a specific
>> residual structure (auto-correlation). Based on my understanding of what
>> nlme and lme4 do well, I would normally turn to lme4 to fit a model with
>> crossed random effects, but because I'm trying to structure the residuals,
>> I am trying nlme.
>> In trying to fit and compare the same variance components (no fixed
>> effects) model using lme4 and nlme, I found the output is similar but a
>> different. Specifically, the standard deviations of the random effects and
>> the log-likelihood statistics are different. Would you expect the output
>> be a bit different?
>> The models I fit to compare the output are here, though the output is also
>> m_lme4 <- lmer(diameter ~ 1 + (1 | plate) + (1 | sample), data =
>> m_nlme <- lme(diameter ~ 1, random = list(plate = ~ 1, sample = ~ 1), data
>> = Penicillin)
>> Thank you for considering this question,
>> Joshua Rosenberg, Ph.D. Candidate
>> Educational Psychology
>> Educational Technology
>> Michigan State University
>> [[alternative HTML version deleted]]
>> R-sig-mixed-models at r-project.org mailing list
Joshua Rosenberg, Ph.D. Candidate
Michigan State University
[[alternative HTML version deleted]]
More information about the R-sig-mixed-models