Online Resource 1: R Code of Model Selection

Online Resource 1: R code of model selection

Manuscript title

Effects of two contrasting hemiparasitic plant species on biomass production and nitrogen availability

Authors

Andreas Demey, Els Ameloot, Jeroen Staelens, An De Schrijver, Gorik Verstraeten, Pascal Boeckx, Martin Hermy, Kris Verheyen

Address for correspondence

Forest and Nature Lab, Ghent University, Geraardsbergsesteenweg 267, B-9090 Gontrode, Belgium

Tel: +32 (0)9 264 90 35

First, we optimized the random structure of our model by comparing models with full fixed effects and different random structure using a Restricted Maximum Likelihood (REML) approach (Zuur et al., 2009). The model with the lowest Akaike Information Criterium (AIC) was retained. The weeding model was compared with an intercept-only model and the weeding*year model was compared with the weeding + year (without interaction effect) model using likelihood ratio tests. The fit of the final model was then checked by assessing normality and homogeneity using graphical tools.

UEffect of weeding of the hemiparasite on aboveground biomass in year 2010

FULL MODEL: biomass ~ 1 + weeding + (1 | location/block)

#STEP 1: Selection of the random structure

>M1 <- lmer(biomass ~ 1 + weeding + (1 | location), REML=T, data)

>M2 <- lmer(biomass ~ 1 + weeding + (1 | location/block), REML=T, data)

>anova(M1, M2)

#STEP 2: Testing for the weeding effect (random structure depending on previous step)

M10 <- lmer(biomass ~ 1 + weeding + (1 | location), REML=T, data)

M11 <- lmer(biomass ~ 1 + (1 | location), REML=T, data)

anova(M10,M11)

UEffect of year:weeding interaction on aboveground biomass for 2009 and 2010 data combined

FULL MODEL: biomass ~ 1 + weeding + year + weeding:year + (year-1| location/block) + (1 | plot)

#STEP 1: Selection of the random structure

M1 <- lmer(biomass ~ weeding + year + weeding:year+(1 | plot), REML=T, data)

M2 <- lmer(biomass ~ weeding + year + weeding:year+(year-1|location)+(1 | plot), REML=T, data)

M3 <- lmer(biomass ~ weeding + year + weeding:year+(year-1|location/block)+(1|plot),REML=T, data)

anova(M1, M2, M3)

#STEP 2: Testing for the weeding effect (random structure depending on previous step)

M30 <- lmer(biomass ~ weeding + year + weeding:year+(year-1 | location)+(1 | plot), REML=F, data)

M31 <- lmer(biomass ~ weeding + year + (year-1 | location)+(1 | plot), REML=F, data)

anova(M30, M31)

Reference

Zuur, A.F., Ieno, E.N., Walker, N.J., Saveliev, A.A., Smith, G.M., 2009. Mixed Effects Models and Extensions in Ecology with R. Springer, New York.