0
$\begingroup$

I would like to apply MDS to a high-dimensional distance matrix but the difficulty is than the distance matrix contains many missing and censored values (i.e. distances like >8).

Does anyone know of any MDS algorithms that can cope with this kind of data or have any advice for what they would do in this type of situation?

$\endgroup$
  • $\begingroup$ If I remember correctly mds in spss (proxscal command) has something like that. $\endgroup$ – ttnphns Jun 30 '19 at 20:46
  • $\begingroup$ Thanks a lot for the pointer, it looks like PROXSCAL certainly deals with missing values, but I can't work out how to specify the censored distances. I've never used SPSS before and could well be missing something? $\endgroup$ – Sam Jul 10 '19 at 13:48
  • $\begingroup$ Well, hmm, I don't know if it can deal with imprecise (censored) distances. I expect you should manually set some value for them. Note that nonmetric MDS greets only order of distance values so it should appear comparatively robust in case of large censored distances. $\endgroup$ – ttnphns Jul 10 '19 at 15:10

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.