0
$\begingroup$

I am analyzing data on stream community before and after an invasive species removal event and have a dataset of dissimilarity values comparing the stream community at each time point in the dataset with the average community pre-removal. I have constructed a model for this dataset dissimilarity ~ cycle + discharge + biomass where cycle is the time of the measurement, discharge is mean stream discharge, and biomass is the biomass of invasive species removed from the stream as a measure of initial invasive community. Each variable is significant, as are the interactions cycle:discharge and cycle:biomass.

I am trying to assess the persistence of removal effects across discharge and biomass. I have the following code that will produce a graph very near what I want using the glht function in the multcomp library.

## create a graph of difference in predicted dissimilarity between cycle 8.2
##  (first post removal time point) with cycle 1
# create a sequence of discharge values
Mean.Q.seq<-seq(0.05,0.7,0.05)
# create a matrix for glht function providing coefficients for the model
#  components (i.e. 0s for unused components, 1*adjustment of the
#  intercept for cycle 8.2, Mean.Q.seq*adjustment of beta Mean.Q:Cycle8.2,
#  average biomass of 4750.5*adjustment of beta Inv.biomass:Cycle8.2)
K.1<-cbind(0,0,0,1,0,0,0,0,0,0,0,0,0,0,0,0,Mean.Q.seq,0,0,0,0,0,0,0,0,0,0,0,0,4750.5,0,0,0,0,0,0,0,0,0,0,0,0)
rownames(K.1)<-paste("Mean.Q",Mean.Q.seq)
# use glht to calculate the differences
bycycle.for.diffs<-glht(bycycle.model.0.b,linfct=K.1)
# calculate the confidence intervals at each discharge value
bycycle.ci<-confint(bycycle.for.diffs,level=0.95)
# create graph
par(mfrow=c(1,1))
plot(Mean.Q.seq,coef(bycycle.for.diffs),type="b",
     ylim=c(min(bycycle.ci$confint[,"lwr"]),max(bycycle.ci$confint[,"upr"])),
     xlab="Mean discharge",ylab="Predicted Dissimilarity, Cycle 8.1 - Pre-removal average")
lines(Mean.Q.seq,bycycle.ci$confint[,"upr"])
lines(Mean.Q.seq,bycycle.ci$confint[,"lwr"])
abline(h=0,lty=3)

My question is whether there is any way to create this graph where the y values are the difference between predicted dissimilarity at cycle 8.2 and the average predicted dissimilarity for pre-removal time points (cycles 1-8.1) such that I can show with the CIs at what range of flow values the community is different from the pre-removal average at this time point? Comparing it to cycle 1 assumes that dissimilarity at cycle 1 is representative of general dissimilarity pre-removal, which is not given.

$\endgroup$

Your Answer

By clicking "Post Your Answer", you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.