[R-sig-ME] grasshoppers and mixed models

Frank Dziock frank.dziock at tu-berlin.de
Wed Jan 28 17:33:02 CET 2009

Hi there!

our response variable is the number of grasshoppers (abundance) on ski
slopes/non ski study plots in alpine ecosystems.

We are interested in the fixed effects of skiing, fertilization,
management intensity and other properties of the study plots on
grasshopper abundance.

We have 82 study plots arranged as pairs (one plot on a ski slope, the
other away from the slope), we call this a BLOCK. We studied these plots
in three different study areas. This lead us to believe we have a nested
design with random effect= 1~AREA/BLOCK

data table arranged as follows:

AREA	BLOCK	grasshopper	fertilizer	management	skiing
1	0	8		1		2		0
1	1	4		1		1		1
1	0	16		0		3		0
1	1	3		1		0		1
2	0	4		0		2		0
2	1	5		1		2		1


The number of grasshoppers recorded is a count variable, so we used

model1 <- glmmPQL(abundance~fertilizer+management,random=~1| AREA/BLOCK,

I was wondering, whether a GLMM like this would be appropriate to decide
which predictor variables have distinct effects on grasshopper numbers.
We are especially unsure, whether the random effect has been properly
defined, because one of the predictor variables (skiing) is also our
BLOCK factor.

We have 15 predictor variables in total and in order not to overfit our
model we did the following:

1. Calculate separate models for single predictor variables (linear,
quadratic and logarithmic terms)
2. We included only the most significant terms of each variable in the
initial model
3. Model was simplified by stepwise removing variables according to
their p-values
4. This was done until all remaining varibles had p-values below 0.05

As I followed the former discussions in your group on p-values, you will
probably want to slaughter me for that approach. But I am an absolute
beginner in mixed models and it would be very helpful for me to receive
a hint, how I could decide which variables play a major role in
determining my grasshoppers abundance while taking into account the
nested study design structure.

Comments would be very welcome!!!

Best wishes from cold Berlin,



Prof. Dr. Frank Dziock
Animal Ecologist and Head of Department (Juniorprofessor)

Department of Biodiversity Dynamics
Technische Universitaet Berlin
Sekr. AB 1
Rothenburgstr. 12
D - 12165 Berlin

Tel: 030 – 314 71368

Secretary Gisela Falk
gisela.falk at tu-berlin.de
Tel: 030 - 314 71350
Fax: 030 – 314 71355


More information about the R-sig-mixed-models mailing list