Hi Natalia,
if your aim is to do ordinations of groups (i.e.,
scatterplots), the BG-PCA might be a very good
idea. However, it's not a classification method (like DA/CVA).
The easiest option I know to do a BG-PCA is using
PAST: menu multivariate, PCA, option between
groups (at least in version 2; it will be similar I guess in version 3).
There's a brief intro on BG-PCA in Seetah et al.
(my webpage - publications) and something on
DA/CVA in Kovarovic et al. (same page). In those
papers you'll find plenty of refs to much more
rigorous work on these topics and in Kovarovic we
say something also on sampling (with refs).
There's more on DA/CVA also in Evin et al.
Besides possible issues with degrees of freedom
in some analyses if N is very small, having very
heterogeneous sample sizes across groups tend to
make things worse and almost inevitably leads to
poor estimates of mean shapes, variances etc. For
instance, in Fig. 7 of Cardini et al., 2015 (on
sampling error), you'll easily see how, when N is
reduced, the error in mean shapes and
multivariate variance in a given population
become massively large compared to differences in
those study samples. If at least one of your
sample is large, you could try a similar experiment and see what happens.
Also, whether you can get away with small N may
depend on the magnitude of the differences among
groups: to get robust results, if you compare
geographic populations of the same species,
you're more likely to need large N than if you're comparing genera.
Sometimes one has to acknowledge the limits of
sampling. There might still be options for
interesting exploratory analyses and one could
flag analyses which involve small samples and
warn that those results will require a confirmation.
Good luck.
Cheers
Andrea
At 14:12 06/02/2015, Natalia Siomava wrote:
Hello everybody,
I am not very experienced in morphometrics but I
need to do some analysis for fly wings. I want
to compare 6 groups. The problem is that for
some groups the sample size is very small - 5
flies, that means after averaging left-right
sides 5 wings. As I understand, I cannot do CVA.
I found, that it's possible to perform
between-group PCA in this case. Has anybody done
it before? Could you explain how to do it?
By the way, how big should be the sample to do DFA?
Thanks in advance.
--
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].
Dr. Andrea Cardini
Researcher in Animal Biology, Dipartimento di
Scienze Chimiche e Geologiche, Università di
Modena e Reggio Emilia, l.go S. Eufemia 19, 41121 Modena, Italy
Adjunct Associate Professor, Centre for Forensic
Science , The University of Western Australia, 35
Stirling Highway, Crawley WA 6009, Australia
E-mail address: [email protected], [email protected]
WEBPAGE: http://sites.google.com/site/hymsfme/drandreacardini
Summary of research interests at:
http://www.dscg.unimore.it/site/home/ricerca/aree-di-ricerca/evolution-taxonomy-and-forensics.html
FREE Yellow BOOK on Geometric Morphometrics:
http://www.italian-journal-of-mammalogy.it/issue/view/405
or full volume at:
http://www.italian-journal-of-mammalogy.it/public/journals/3/issue_241_complete_100.pdf
Editorial board for:
Zoomorphology:
http://www.springer.com/life+sciences/animal+sciences/journal/435
Journal of Zoological Systematics and
Evolutionary Research: http://www.wiley.com/bw/journal.asp?ref=0947-5745&site=1
Hystrix, the Italian Journal of
Mammalogy: http://www.italian-journal-of-mammalogy.it/
--
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].
