0
$\begingroup$

I have a set of dummy variables (~300) indicating a particular feature, and rows which represent an individual.

I plot this data after using nMDS to visualize which individuals are more similar to other individuals. And this seems to work as I expect.

However, now I want to quantitatively explain why the space is the way it is. I.e. why aren't the points randomly distributed across the space. As I understand from this post, I shouldn't try to model the nMDS axis (although I could possible use the dimensions to explain a response?)

I have thought that perhaps I could perform a clustering analysis on the outputs, and then model the classifications using some modelling techniques (random forest, categorical linear models or whatever). Does this seem appropriate? Are there other methods that might be more appropriate for this?

$\endgroup$
  • $\begingroup$ MDS is an analysis of a distance matrix, it starts with the square symmetric distance matrix between objects (individuals). What was your distance measure? $\endgroup$ – ttnphns Aug 13 at 8:52
  • $\begingroup$ The data is converted to a distance matrix using dist(., "binary") function in R. Although I have also just used a convenience function metaMDS() which does this internally. $\endgroup$ – SamPassmore Aug 13 at 8:57
  • $\begingroup$ Just a remark: Please mind that this site is not a software site, but a statistical one. Here is a lot people who don't understand what is "dist(., "binary") function in R" or "metaMDS()". $\endgroup$ – ttnphns Aug 13 at 11:16
  • $\begingroup$ R is the most commonly used question tag on this site so I disagree on this point. I can provide more detail if you think it will be useful - but I don't understand why it helps answer the question here. Perhaps you can explain why the distance measure is important for this? $\endgroup$ – SamPassmore Aug 13 at 11:42

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.