How to obtain procustes distances among groups

802 views
Skip to first unread message

Madlen Stange

unread,
Oct 3, 2016, 9:32:23 AM10/3/16
to geomorph R package

Dear geomorph users,

I'm wondering how I can calculate Procrustes distances between groups (e.g. species) of samples. In MorphoJ this is calculated after performing a CVA.

The output of PV.dist of morphol.disparity() are distances in some sense, but are they distances among centroids of groups? What me confuses is the term absolute differences between variances in the help page of the command. I'm not sure what this means in this context. Can I use the morphol.disparity output table and related p-values in order to make statements about how disparate groups are from each other or how similar they are?

If not, what else is commonly used to quantify similarity/dissimilarity among groups?

Thank you very much for your help in advance!
Madlen

Mike Collyer

unread,
Oct 3, 2016, 9:54:35 AM10/3/16
to Madlen Stange, geomorph R package
Madlen,

You do not want to use mrophol.disparity for that.  That function compares Procrustes variances.  The distances between group centroids, along with P-values is output from advanced.procD.lm.  The help file has some instructions for its use.

Please note that unlike the CVA option in morphoJ, one can define different null models with advanced.procD.lm.  The equivalent null model for the one used in CVA in morphoJ is ~1, I believe.  

Cheers!
Mike


--
You received this message because you are subscribed to the Google Groups "geomorph R package" group.
To unsubscribe from this group and stop receiving emails from it, send an email to geomorph-r-pack...@googlegroups.com.
To post to this group, send email to geomorph-...@googlegroups.com.
Visit this group at https://groups.google.com/group/geomorph-r-package.
To view this discussion on the web, visit https://groups.google.com/d/msgid/geomorph-r-package/e1b2d709-72d5-4d13-b18f-0a973ad83583%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

Madlen Stange

unread,
Oct 3, 2016, 11:12:56 AM10/3/16
to geomorph R package
Thank you very much for your reply, Michael!

I managed to gain the same results for group distances with advanced.procD.lm as when I computed them handish. I was not sure I'm doing it right.

Ok, so does the morphol.disparity output rather mean how much more or less disparate, expressed in procrustes variance, a defined group is compared to the other group? Say, group X vs. Y yields 0.003 and X vs. Z yields 0.03, does that mean X and Y are more similar in variance than X and Z?
And in general, can variance be translated into space they occupy in morphospace?

Best,
Madlen

Mike Collyer

unread,
Oct 3, 2016, 11:20:35 AM10/3/16
to Madlen Stange, geomorph R package
It was pointed out to me that the shapes package has greater flexibility for calculating different types of distances (e.g., full and partial Procrustes, Riemannian), if that is of interest.  It should be noted that the distances in advanced.procD.lm are Euclidean distances, which would mimic Procrustes distances following projection into tangent space (assuming projection was used with GPA).

Cheers!
Mike

Mike Collyer

unread,
Oct 3, 2016, 11:39:07 AM10/3/16
to Madlen Stange, geomorph R package
You are welcome, 

And the answer is”yes” to both of your questions.

Cheers,
Mike


--
You received this message because you are subscribed to the Google Groups "geomorph R package" group.
To unsubscribe from this group and stop receiving emails from it, send an email to geomorph-r-pack...@googlegroups.com.
To post to this group, send email to geomorph-...@googlegroups.com.
Visit this group at https://groups.google.com/group/geomorph-r-package.
Reply all
Reply to author
Forward
0 new messages