axis break in plot metaMDS

Question

I have an abundance table of fungal species from 16 different samples (xylem tree tissue) belonging to three healthy classes.

I want to visualize the similarity of these samples according to the health class. I have run metaMDS (vegan package) and plot the output from metaMDS, here's an example of the distances obtained (data_mds)

      NMDS1        NMDS2
1  -25.82806 -0.216970172
2  -24.59347 -0.377391024
3  -25.80740  0.268355050
4  -23.41391 -0.793553278
5  -25.93017  0.179795622
6  -25.39369  0.471681826
7  -25.29794  0.044754740
8  -24.68337  1.612166365
9  400.26576  0.003309978
10 -23.21608 -0.632045558
11 -22.67440  0.268931564
12 -23.93604  0.551203963
13 -23.55546 -1.304642023
14 -25.77035 -0.989877602
15 -25.77835 -0.226395569
16 -24.38708  1.140676118

If I plot them, sample 9 will skew all the graph, so all the other 15 samples will overlap in one place (as you can see in the figure).

15 samples overlapping and sample no.9 alone

I have tried using gap.plot to create a gap between the two groups but I can't figure out how to create a plot where the 15 samples are more separated (both x and y axes) and then having a separate part for sample 9. With this code I have manage to create two separate plots but as you can see in figure 2 the 15 samples are not spread out on the x axis.

nmds_plot <- plot(data_mds, main = "NMDS", type="none")
gap.plot(nmds_plot, gap=c(300, 350), gap.axis="x", xlim=c(-30, 500), ylim = c(-2, 2))

cols <- c("darkturquoise","darkturquoise","darkturquoise","darkturquoise","darkturquoise","darkturquoise","darkturquoise","darkturquoise",
      "grey0","grey0","grey0","grey0",
      "chocolate1","chocolate1","chocolate1","chocolate1")
pch_data_mds<-c(15,15,15,15,15,15,15,15,17,17,17,17,19,19,19,19)

points(nmds_plot, col=cols, pch=pch_data_mds, cex=1.4)

abline(v=0, col="gray59", lty="dotted")
abline(h=0, col="gray59", lty="dotted")

colsleg <- c("darkturquoise", "grey0", "chocolate1")
legend("topright", legend = c("Low", "Medium", "High"), col=colsleg , pch=c(15,17,19), cex=0.5, title = "Vitality classes")

graphs divided but with the 15 samples not spaced on the x axis

Thank you for the help

Jari Oksanen Jari Oksanen · Accepted Answer · 2017-06-08T04:31:38

You have discontinuity and an outlier. There is no way of having axis breaks in ordination diagrams, but the result is what it is. However, you can have xlimand ylim arguments to limit the plot to the desired range. Usually you have to give both limits, because the ordination plot will have equal aspect ratio. However, often you find little structure in the overlaid bunch of points, but you need to deal with the outlier. Often the reason to be is an outlier is to have nothing in common with the rest of the points. Removing an outlier often helps. Sometimes it helps to set noshare = TRUE in the metaMDS call.

axis break in plot metaMDS

1 Answers