I have a vague recollection from some years ago of lda() having a small
bug/feature. Something like division by n instead of (n-1) or
vice-versa. As always, best to check code and results against other,
independent programs and resolve the reason for any differences. -ds
On 8/19/15 7:08 PM, Fábio Machado wrote:
I would consider using the lda() function from MASS package. The
function allows for the use of a training dataset (the extant data, in
your case) and the associated predict function gives you the score of
all specimens (extant and fossil). In my experience, the ordination of
specimens on the discriminant functions of lda and on the canonical
variate axis of cva from Morpho are identical, except for arbitrary
inversion of the sign of some axis, usually the first. Additionally,
predict.lda already give you group membership probabilities.
As for PCA, I think that princomp and prcomp functions have a predict
option that would produce scores for specimens not included in the
original analysis.
Best,
Fabio Andrade Machado
Laboratório de Evolução de Mamíferos
Departamento de Genética e Biologia Evolutiva- USP
[email protected] <mailto:[email protected]> ; [email protected]
<mailto:[email protected]>
+55 11 982631029
skype: fabio_a_machado
Lattes: http://lattes.cnpq.br/3673327633303737
Google Scholar: http://scholar.google.com/citations?hl=en&user=2l6-VrQAAAAJ
Em 19/08/2015, à(s) 18:40, Blake Dickson <[email protected]
<mailto:[email protected]>> escreveu:
Hey Morphmetricians,
So I have a question regarding calculating PC scores and CV scores
post-hoc:
I have a GMM dataset of extant and fossil species which I want to plot
using PCA and CVA. I want to calculate the extant species first, then
use the eigenvectors from this the plot the scores for the fossil taxa.
I have done this successfully for the PCA in R by centering the whole
dataset, then calculating the covariance matrix and eigenvectors for
the extant species only. I then calculate the PC scores for the fossil
taxa using these eigenvalues.
I am not certain whether the same tactic for the CVA is valid. I have
tried it: performing the CVA (using the Morpho package) on the extant
dataset with associated groupings, then correcting the fossil
coordinates by the mean of this CVA and calculating the CV scores for
the fossil taxa using the canonical variates (CV) matrix. This gives
me a result, with plot-able CV scores for the fossil taxa, but I want
to be certain the method is valid.
In addition, assuming that what I have done with the CVA is correct;
how would I best go about testing the likelihood of group membership
for these fossil taxa?
Cheers all,
Blake Dickson
--
MORPHMET may be accessed via its webpage at
http://www.morphometrics.org <http://www.morphometrics.org/>
To unsubscribe from this group and stop receiving emails from it, send
an email to [email protected]
<mailto:[email protected]>.
--
MORPHMET may be accessed via its webpage at http://www.morphometrics.org
--
MORPHMET may be accessed via its webpage at http://www.morphometrics.org
To unsubscribe from this group and stop receiving emails from it, send an email
to [email protected].
