As a beginner to MDS, here is my thought process:

Given a data set of environmental factors that may effect a certain sites, when I run a PCA on each site I get a list of principal components. If I want to understand which environmental factors are driving each of the principal components, ie:

PCA1 <- Mostly Oxygen and Salinity driven
PCA2 <- Mostly Temperature driven

I look at the loadings of the PCA results to find out which of the original environmental factors are weighting each of the principal components.

When I run a NMMDS on a similar set of data, my result is a plot in which the sites are grouped so that more dissimilar sites are further apart. While I can make an educated guess as to what is causing the grouping in an MSD plot, is there something analogous to the loadings of a PCA that explains which environmental variables may be playing the largest roles in the resulting plot? Or a way to quantify the input of each environmental variable to the final structure of the plot?

  • $\begingroup$ Loadings are (regressional) weights of components on variables, not of variables on components. $\endgroup$ – ttnphns Dec 2 '13 at 19:01
  • $\begingroup$ Please, tell more about how your data look. E.g.: when you do PCA on columns of data, what define rows and what define are columns? When you do MDS, what define you dissimilarity matrix and what dissimilarity you use? $\endgroup$ – ttnphns Dec 2 '13 at 19:07
  • $\begingroup$ I don't have any specific answers for my data yet, still in the nascent stages. This is more a general question to get an idea of what is possible with NMDS, and will I be able to quantify some of the visual results I see in my NMDS plots $\endgroup$ – Vinterwoo Dec 3 '13 at 18:44

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.