Non-Metric Multidimensional Scaling

Non-Metric Multidimensional Scaling (NMDS) is a distance-based ordination that focuses on preserving the rank order of pairwise dissimilarities.

Description

Non-Metric Multidimensional Scaling (NMDS) is a distance-based ordination commonly used in ecology and other fields to visualise and explore the (dis)similarity relationships between a set of observations. Unlike PCOA, NMDS does not rely on preserving the actual distances between points, but rather focuses on preserving the rank order of pairwise dissimilarities. Another property of NMDS is that the axes are orthogonal (no difference in eigenvalues) and can therefore be freely rotated.

While this ordination is not sensitive to double-zero effects, it can take a long time to compute on large datasets.

Example

lalashan <- readRDS("lalashan.rds")

library(vegan)

nmds_ord <- metaMDS(comm = lalashan$cross, distance = "bray")

plot(nmds_ord)

stressplot(nmds_ord)

Alternative functions

• vegan::metaMDS()
• MASS::isoMDS()

Further References

• https://www.davidzeleny.net/anadat-r/doku.php/en:pcoa_nmds
• https://en.wikipedia.org/wiki/Multidimensional_scaling