[R-sig-ME] glmer with dissimilarity values

[Douglas Bates]

Generally it is best to send questions like this to the
R-SIG-Mixed-Models at R-project.org mailing list (see instructions at
https://stat.ethz.ch/mailman/listinfo/r-sig-mixed-models).  Many of
those who read that list can reply and often much faster than I am
able to.


On Mon, Oct 22, 2012 at 4:22 AM, Eva Pitta <epitta at upatras.gr> wrote:
> Hello,
> I am a phd student in Greece using the lme4 package to fit generalized
> linear mixed-effects models.
> I have a question regarding the use of glmm on certain type of data.
> The problem is that my data are dissimilarity values calculated using the
> one-complement of the Jaccard index
> for each pair of islands in each archipelago. If an archipelago has 10
> islands I end up with 45 dissimilarity values.
> If a second archipelago has 5 islands I end up with 10 dissimilarity values.
> I know that dissimilartiy matrices are analysed using Mantel tests or
> multiple regression on distance matrices.
> The data are first in the form of a dissimilarity matrix for each
> archipelago. For each archipelago I used Mantel test to correlate the
> dissimilarity matrix with the geographical distance matrix (distance was
> calculated between all islands of an archipelago).
> However, this approach (or multiple regression on distance matrices) cannot
> be applied to many
> dissimilarity matrices from different areas (or archipelagos as is my case)
> at the same time.
> So I created a vector with the dissimilarity values between all island pairs
> and also another vector with the distances between the islands and
> I fitted a glmm with dissimilarity as a response variable, distance as a
> fixed effect and archipelago as random effect. I also included island type
> as a fixed effect beacuse I want to see if there are differences between
> oceanic and continental islands.
> I know that glmms allow for correlation betwen observations. Does this mean
> that I can use glmm to model dissimilarity values?
> Dissimilarity values from different archipelagos are independent from each
> other. I only calculate dissimilarity values between islands of the same
> archipelago.
>
> For example my data are in the form:
>
> Dissimilarity   Distance        Island_type     Archipelago
> 0               5259.157        continental     adriatic_islands
> 0.4285714       8967.4931       continental     adriatic_islands
> 0.5714286       261975.283      continental     adriatic_islands
> 0.5714286       25605.9319      continental     adriatic_islands
> 0.5714286       85230.7187      continental     adriatic_islands
> 0.7142857       298784.1031     continental     adriatic_islands
> 0.7142857       369245.918      continental     adriatic_islands
> 0.7142857       251044.0615     continental     adriatic_islands
> 0.8571429       231761.9685        continental     adriatic_islands
> 0.8571429       58223.7633      continental     adriatic_islands
> 0.8571429       253342.5877        continental     adriatic_islands
> 0.8571429       329540.3376     continental     adriatic_islands
> 0.4285714       20047.0762      continental     adriatic_islands
> 0.5714286       248763.7943        continental     adriatic_islands
> 0.5714286       21853.6469      continental     adriatic_islands
> 0.3             89478.116       continental     central_aegean_islands
> 0.5454545       74230.855       continental     central_aegean_islands
> 0.5454545       125774.88       continental     central_aegean_islands
> 0.5454545       155256.578      continental     central_aegean_islands
> 0.5454545       126187.824      continental     central_aegean_islands
> 0.5             180264.809      continental     central_aegean_islands
> 0.6363636       180998.609      continental     central_aegean_islands
> 0.6363636       147414.741      continental     central_aegean_islands
> 0.75            146702.402      continental     central_aegean_islands
> 0.6363636       174237.948      continental     central_aegean_islands
> 0.6             61902.301       continental     central_aegean_islands
> 0.6             48277.908       continental     central_aegean_islands
> 0.7272727       103213.926      continental     central_aegean_islands
> 0.7272727       134316.409      continental     central_aegean_islands
> 0.7272727       169854.705      continental     central_aegean_islands
> 0.5384615       542269.8027     oceanic         solomon_islands
> 0.5             523544.4758     oceanic         solomon_islands
> 0.5833333       8815.5537       oceanic         solomon_islands
> 0.5833333       276307.3533        oceanic         solomon_islands
> 0.6666667       111280.2721     oceanic         solomon_islands
> 0.5454545       211541.7332     oceanic         solomon_islands
> 0.6666667       706676.0145        oceanic         solomon_islands
> 0.6363636       143283.6463     oceanic         solomon_islands
> 0.75            44164.1395      oceanic         solomon_islands
> 0.7272727       293562.926      oceanic         solomon_islands
> 0.7272727       168887.5203     oceanic         solomon_islands
> 0.8333333       167874.4115     oceanic         solomon_islands
> 0.8181818       1273543.673     oceanic         solomon_islands
> 0.8181818       747544.6154        oceanic         solomon_islands
> 0.8181818       270851.9093        oceanic         solomon_islands
> 0.8181818       1178250.474     oceanic         solomon_islands
> 0.8181818       853046.0902     oceanic         solomon_islands
> ...
>
> The non-independence of dissimilarity values of the same archipelago is due
> to the fact that they are calculated using partly the same data. Is this is
> accounted for by the random factor?
> I would appreciate any help you can give me.
>
> Thank you for your time
>
> Eva Pitta
>