##Correspondence Analysis
library(vegan)
speabu<-read.csv('abund.csv', header=T, row.names=1)
#transform to presence absence (binary)
speocc<-data.trans(speabu[,-1], method='power', exp=0, plot=F)
#perform CA
spe.ca<-cca(speocc)
#results are complex and sotred in a list that gives eigenvalues (intertia), and eigenvectors(sample and species scores)
summary(spe.ca)
#use a randomization test to look at statistical significance of the first several eigenvalues
ordi.monte(speocc[,-1],ord='ca', dim=5, perm=500)
# we can also look at this info graphically
ordi.scree(spe.ca,ord='ca')
#to look at the sample and species scores.. ('u' is sample scores, 'v' is species scores)
spe.ca$CA$u[,1:2] 
spe.ca$CA$v[,1:2]
#or look at scaled scores
spe.ca$CA$u.eig[,1:2]
spe.ca$CA$v.eig[,1:2]
#in CA choice of scaling type is important either the ordination of rows (samples) or columns (species) is meaningful and can be interpreted as an approximation of the chi-square distances between samples or species, respectively
#in scaling type 1: sample scores are calculated as weighted averages of species scores, in scaling type 2: species scores are calculated as weighted averages of sample scores, scaling type 3: sample and species scores are scaled symmetrically (both by the square root of the eigenvalues)
#create a CA joint plot
ordiplot(spe.ca, choices=c(1,2), scaling=2)
#or a labelier one
ordiplot(spe.ca, type='n', scaling=2)
text(spe.ca$CA$v.eig[,1:2],labels=row.names(spe.ca$CA$v.eig), cex=0.7)
text(spe.ca$CA$u[,1:2], labels=row.names(spe.ca$CA$u), cex=0.7, col='red')
#DETRENDING - we didn't really like this in class so I'll write it out but probably won't use it.  Seems fakey and manipulative (but done to get rid of horseshoe effect)
spe.dca<-decorana(speocc,ira=0)
spe.dca
summary(spe.dca)
plot(spe.dca)
#going to try this with my catagorical data
exonFC<-read.csv('exonsforconting.csv', header=TRUE, row.names=1)
summary(exonFC)
exon.ca<-cca(exonFC)
#ok, doesn't work, says it needs to be numeric.  I thought it worked off a contingency table.