Search code examples
rggplot2vegan

ggplot2 version of shepard plot, i.e. vegan::stressplot()?

There seems to be quite a bit of information for plotting NMDS outputs (i.e. NMDS1 vs NMDS1) using ggplot2 however I cannot find a way to plot the vegan::stressplot() (shepard's plot) using ggplot2.

Is there a way to produce a ggplot2 version of a metaMDS output?

Reproducible code

library(vegan)

set.seed(2)

community_matrix = matrix(
  sample(1:100,300,replace=T),nrow=10,
  dimnames=list(paste("community",1:10,sep=""),paste("sp",1:30,sep="")))

example_NMDS=metaMDS(community_matrix, k=2) 

stressplot(example_NMDS)

Created on 2021-09-17 by the reprex package (v2.0.1)

Solution

  • Here's a workaround to plot a very similar plot using ggplot2.The trick was to get the structure of the stressplot(example_NMDS) and extract the data stored in that object. I used the tidyverse package that includes ggplot and other packages such as tidyr that contains the pivot_longer function.

    library(vegan)
library(tidyverse)

# Analyze the structure of the stressplot
# Notice there's an x, y and yf list
str(stressplot(example_NMDS))

# Create a tibble that contains the data from stressplot
df <- tibble(x = stressplot(example_NMDS)$x,
       y = stressplot(example_NMDS)$y,
       yf = stressplot(example_NMDS)$yf) %>%
  # Change data to long format
  pivot_longer(cols = c(y, yf),
               names_to = "var")

# Create plot
df %>%
  ggplot(aes(x = x,
             y = value)) +
  # Add points just for y values
  geom_point(data = df %>%
               filter(var == "y")) +
  # Add line just for yf values
  geom_step(data = df %>%
              filter(var == "yf"),
            col = "red",
            direction = "vh") +
  # Change axis labels
  labs(x = "Observed Dissimilarity", y = "Ordination Distance") +
  # Add bw theme
  theme_bw()

    stressplot