[R-sig-eco] temporal autocorrelation with lme

[Thierry Onkelinx]

Dear Jean-Yves,

The mailing strips HTML and most of the attachment. Hence we can't see the
plots.

Note that you need the normalised residuals to see the effect of the
correlation structure. resid(type = "normalized").

Best regards,

ir. Thierry Onkelinx
Instituut voor natuur- en bosonderzoek / Research Institute for Nature and
Forest
team Biometrie & Kwaliteitszorg / team Biometrics & Quality Assurance
Kliniekstraat 25
1070 Anderlecht
Belgium

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

2016-04-07 14:13 GMT+02:00 Jean-Yves BARNAGAUD <
jean-yves.barnagaud at cefe.cnrs.fr>:

> Dear all,
>
> I'm trying to fit lme models on spatial-temporal data with strong
> inter-annual autocorrelation.
> The response variable is a bird diversity measure sampled on several plots
> during 10 to 40 years, which I want to relate to environmental covariates.
> To give an idea of the sample size, there are appr. 800 plots.
>
> Here is the temporal autocorrelation of the response variable (averaged
> over years, x axis=years)
>
>
> Based on AIC, the best model I tried is a model with a corAR1 structure,
> as compared with ARMA models with different p and q lags.
>
> x1=corAR1(form=~year | plot)
> m2=lme(y~cov1+cov2+cov3,random=list(plot=~1),correlation=x1,data=d1)
>
> Below are residual plots. The first plot is a plot of residuals over
> years, the 2nd and third are plots of mean residuals per year vs years. The
> third plot is from the acf function. Looks like the AR structure does not
> correct temporal autocorrelation, and the ARMA models don't do better. I've
> tried other things, including adding a "year" covariate to remove the
> apparent linear trend, without making it with a better correction.
>
>
> Any advice or alternative solution?
>
> Many thanks,
>
> Best
>
> JYB
>
>
>
>
> --
> Jean-Yves Barnagaud
> Vertebrate Biogeography & Ecology group
> EPHE - CEFE - Montpellier, Fr.
> +0033(0)467613326
> http://sites.google.com/site/jybarnagaud
>
>
> _______________________________________________
> R-sig-ecology mailing list
> R-sig-ecology at r-project.org
> https://stat.ethz.ch/mailman/listinfo/r-sig-ecology
>

	[[alternative HTML version deleted]]