Customizing a vegan pca plot with ggplot2 - r

Question

I'm trying to make a custom plot of some vegan rda results in ggplot2. I'm essentially modifying directions as seen in Plotting RDA (vegan) in ggplot, so that I am using shape and color labels to convey some information about the sample points.
I set up a pca analysis with vegan as follows
library(vegan)
library(dplyr)
library(tibble)
library(ggplot2)
cbPalette <- c("#000000", "#E69F00", "#56B4E9", "#009E73", "#0072B2", "#D55E00", "#CC79A7", "#F0E442")
data(dune)
data(dune.env)
dune.pca <- rda(dune)
uscores <- data.frame(dune.pca$CA$u)
uscores1 <- inner_join(rownames_to_column(dune.env), rownames_to_column(data.frame(uscores)), type = "right", by = "rowname")
vscores <- data.frame(dune.pca$CA$v)
I can make a simple biplot
biplot(dune.pca)
Now, lets say I want to know more about the management conditions to which these different samples were subject. I'll color and shape code them and plot with ggplot.
p1 <- ggplot(uscores1, aes(x = PC1, y = PC2, col = Management,
shape = Management)) +
geom_point() +
scale_color_manual(values=cbPalette) +
scale_fill_manual(values=cbPalette) +
scale_shape_manual(values = c(21:25)) +
theme_bw() +
theme(strip.text.y = element_text(angle = 0))
p1
Next, I'd really like to add some biplot arrows that show us the axes corresponding to species abundance. I can use ggplot to plot just those arrows as follows:
p2 <- ggplot() + geom_text(data = vscores, aes(x = PC1, y = PC2, label = rownames(vscores)), col = 'red') +
geom_segment(data = vscores, aes(x = 0, y = 0, xend = PC1, yend = PC2), arrow=arrow(length=unit(0.2,"cm")),
alpha = 0.75, color = 'darkred')
p2
What I'd really like to do though, is get those arrows and points on the same plot. Currently this is the code that I am trying to use:
p3 <- p1 + geom_text(data = vscores, aes(x = PC1, y = PC2, label = rownames(vscores)), col = 'red') +
geom_segment(data = vscores, aes(x = 0, y = 0, xend = PC1, yend = PC2), arrow=arrow(length=unit(0.2,"cm")),
alpha = 0.75, color = 'darkred')
p3
To my annoyance, this yields only a blank plot (empty window, no error messages). Clearly I am missing something or scaling something incorrectly. Any suggestions about how I can best superimpose the last two plots?

Answer 1

Check ggvegan package from github. It is still in 0.0 versions, and not actively developed at the moment, but if you say
library(ggvegan)
autoplot(dune.pca) # your result object
You get this graph which you can customize in the usual ggplot2 way with various aesthetics.

Answer 2

Try:
library(cowplot) #not needed I just had it attached while answering the question hence the theme.
library(ggplot2)
ggplot(uscores1) +
geom_point(aes(x = PC1, y = PC2, col = Management,
shape = Management)) +
scale_color_manual(values=cbPalette) +
scale_fill_manual(values=cbPalette) +
scale_shape_manual(values = c(21:25)) +
geom_text(data = vscores, aes(x = PC1, y = PC2, label = rownames(vscores)), col = 'red') +
geom_segment(data = vscores, aes(x = 0, y = 0, xend = PC1, yend = PC2), arrow=arrow(length=unit(0.2,"cm")),
alpha = 0.75, color = 'darkred')+
theme_bw() +
theme(strip.text.y = element_text(angle = 0))
The p1 plot was passing col and shape variable Management to geom_text/geom_segment since they were not defined there but there is no Management column in data = vscores. At least I think so based on the error:
`Error in eval(expr, envir, enclos) : object 'Management' not found`

Answer 3

You should also take a look at ggordiplots on GitHub (https://github.com/jfq3/ggordiplots). It includes the function gg_env__fit which fits environmental vectors to an ordination plot. All of the functions inthe package silently return data frames that you may use to modify the plots anyway you wish. The package includes a vignette on modifying the plots. You can read the vignettes without having to install the package by going to john-quensen.com and looking at the GitHub page.