A good place to start with all GL(M)M questions is Ben Bolker's FAQ which
provides thoughts on many commonly asked questions. In particular
http://bbolker.github.io/mixedmodels-misc/glmmFAQ.html#how-do-i-compute-a-coefficient-of-determination-r2-or-an-analogue-for-glmms
An approach I have used
Hi everyone,
What is the best way to express the goodness of fit for a model obtained by
generalized least squares (using gls function from nlme package)? I don’t want
to compare this model with other models, so AIC is not a good choice in this
case. Is it right to use the approach proposed by
David,
Assuming the distance/similarity measure is non-euclidean (e.g., Bray-Curtis),
see attached pdf document containing a previous post on this subject. This
approach is based on the one that Marti Anderson describes in Primer.
Good luck,
Steve
Stephen Brewer
Principal response curves would be a good way to go if you have balanced
data.
My recent paper may be of interest:
Bird, TLF et al.. (2017) Shrub Encroachment Effects on Habitat
Heterogeneity and Beetle Diversity in a Mediterranean Coastal Dune System. Land
Degrad. Develop., 28: 2553–2562. doi:
Hi,
You also try using principal response curves as suggested in this recent
publication:
dx.doi.org/10.1002/ecs2.2023
Zoltan
2018.01.24. 15:28 keltezéssel, Pedro Pequeno írta:
> Hi,
>
> you could ordinate your observations first (e.g. using NMDS or PCoA), and
> then model the resulting
Hi,
you could ordinate your observations first (e.g. using NMDS or PCoA), and
then model the resulting scores using location and time as predictors (if
you are interested in estimating their independent effects), or using a
repeated-measures Anova or a GLMM with location as random factor to