问题
When I try to make a CCA plot, only 4 out of 25 vectors are displayed.
So, I start by loading my data...
#Load package and Data
library(vegan)
Species <- read.csv("D:/R/Code/TestSpiders.csv", head = TRUE, row.names = 1)
Plants <- read.csv("D:/R/Code/Plants.csv", head = TRUE, row.names = 1)
And then, I make use the following code to make the CCA and plot it.
#Making the CCA
CCA.Plants <- cca(Species, Plants)
plot(CCA.Plants)
Even though my Plants data has much more than 4 columns (it has 25), only 4 vectors are displayed in the resulting CCA graph. These 4 vectors represent the first 4 columns of my Plants dataframe. This does not happen when I use the data used in the CCA example (varespec and varechem). I cannot see any difference in the format of varechem and my Plants data.
This is the result when I input CCA.Plants
Call: cca(X = Species, Y = Plants)
Inertia Proportion Rank
Total 1.0904 1.0000
Constrained 0.4789 0.4392 4
Unconstrained 0.6115 0.5608 15
Inertia is mean squared contingency coefficient
Some constraints were aliased because they were collinear (redundant)
Eigenvalues for constrained axes:
CCA1 CCA2 CCA3 CCA4
0.19122 0.16817 0.06850 0.05101
Eigenvalues for unconstrained axes:
CA1 CA2 CA3 CA4 CA5 CA6 CA7 CA8 CA9 CA10 CA11 CA12
0.16984 0.12046 0.06941 0.05250 0.04174 0.03349 0.02291 0.02206 0.02015 0.01970 0.01504 0.00957
CA13 CA14 CA15
0.00802 0.00428 0.00228
Any input is appreciated.
回答1:
The clue is here:
Some constraints were aliased because they were collinear (redundant)
in other words, the other 21 columns of Plants could be created as a linear combination of the four plant species shown in the plot. Those 21 plants contained no extra information and hence are dropped from the analysis.
I also don't think the approach you are using here is a good one. Why should the spiders respond unimodally to linear gradients of plants?
Instead, I would suggest using co-correspondence analysis (Co-CA)for this type of problem. Either the symmetric version if you don't want either set of species to play the predictors or responses, or the predictive version if you really want to predict spiders using plant species composition.
This method (both versions) is implemented in R via my cocorresp package, which is on CRAN, and is based on the original Matlab routines provided as supplementary information to the paper that introduced Co-CA (ter Braak & Schaffers, 2004).
For example:
## symmetric CoCA
data(beetles)
## log transform the beetle data
beetles <- log1p(beetles)
data(plants)
## fit the model
bp.sym <- coca(beetles ~ ., data = plants, method = "symmetric")
bp.sym
Giving:
Symmetric Co-Correspondence Analysis
Call: symcoca(y = y, x = x, n.axes = n.axes, R0 = weights, symmetric =
symmetric, nam.dat = nam.dat)
Eigenvalues:
COCA 1 COCA 2 COCA 3 COCA 4 COCA 5 COCA 6 COCA 7 COCA 8
0.2534 0.1289 0.0811 0.0741 0.0585 0.0474 0.0373 0.0320
COCA 9 COCA 10 COCA 11 COCA 12 COCA 13 COCA 14 COCA 15 COCA 16
0.0308 0.0233 0.0207 0.0184 0.0172 0.0161 0.0144 0.0118
COCA 17 COCA 18 COCA 19 COCA 20 COCA 21 COCA 22 COCA 23 COCA 24
0.0106 0.0100 0.0087 0.0085 0.0066 0.0063 0.0050 0.0044
COCA 25 COCA 26 COCA 27 COCA 28 COCA 29
0.0043 0.0034 0.0022 0.0010 0.0006
Inertia:
beetles plants
Total: 3.98833 5.757
Explained: 3.97079 5.740
Residual: 0.01754 0.018
and
layout(matrix(1:2, ncol = 2))
plot(bp.sym, which = "response", main = "Beetles")
plot(bp.sym, which = "predictor", main = "Plants")
layout(1)
Note that in this symmetric analysis, neither set of assemblages plays the response or predictor role, but that is how the plot method chooses which to draw, based on which was on the left or the right hand side of the ~ in the formula.
The predictive Co-CA works similarly.
ter Braak, C.J.F and Schaffers, A.P. (2004) Co-Correspondence Analysis: a new ordination method to relate two community compositions. Ecology 85(3), 834–846
来源：https://stackoverflow.com/questions/36041924/r-cca-only-displaying-4-vectors