multivariate testing - Can mvabund be used for a nested experimental design? -
i'm using mvabund , manyglm test differences in species composition in samples taken different treatment plots (experimental treatment plots or control plots). don't wish use adonis because taxa variance increases mean , taxa have low variance.
i have 24 samples in total (rows in dataset samples) have 181 different species recorded (in columns). species ants , samples trees.
my experimental design nested have 3 trees within each plot , plots either treatment or control. account effect of different tree species if possible.
my first question: trees pseudoreplicates within each plot there way account nested design in mvabund?
my second question: when have applied model below, output not give p values each species in univariate tests. why this?
the model shows significant effect of treatment:
mod2 <- manyglm(data.mvabund ~ data$treatment, family="negativebinomial") anova(mod2)
multivariate test: res.df df.diff dev pr(>dev)
(intercept) 23
data$treatment 22 1 361 0.003 **
signif. codes: 0 ‘’ 0.001 ‘’ 0.01 ‘’ 0.05 ‘.’ 0.1 ‘ ’ 1
but when univariate tests on each species there no p value:
anova(mod2, p.uni="adjusted") multivariate test: res.df df.diff dev pr(>dev)
(intercept) 23
data$treatment 22 1 361 0.003 **
signif. codes: 0 ‘’ 0.001 ‘’ 0.01 ‘’ 0.05 ‘.’ 0.1 ‘ ’ 1
univariate tests: dolichoderus.3 dolichoderus.4 dolichoderus.5
dev pr(>dev) dev pr(>dev) dev pr(>dev) (intercept)
dolichoderus.6 dolichoderus.7 dolichoderus.8
dev pr(>dev) dev pr(>dev) dev pr(>dev) (intercept)
dolichoderus.9 dolichoderus.10 dolichoderus.11
dev pr(>dev) dev pr(>dev) dev pr(>dev) (intercept)
dolichoderus.12 dolichoderus.13 tapinoma.1 tapinoma.2 dev pr(>dev) dev pr(>dev) dev pr(>dev) dev (intercept)
tapinoma.3 tapinoma.4 tapinoma.5
pr(>dev) dev pr(>dev) dev pr(>dev) dev pr(>dev) (intercept)
technomyrmex.4 technomyrmex.5 technomyrmex.6
dev pr(>dev) dev pr(>dev) dev pr(>dev)
thanks
steph
Comments
Post a Comment