-------- Original Message --------
Subject: Re: correlation coefficient between two or more matrices
Date: Wed, 29 Jun 2011 13:08:28 -0400
From: Manabu Sakamoto <m.sakam...@bristol.ac.uk>
To: morphmet@morphometrics.org
Thank you very much for everyone who has replied to my question either
on- or off-list. I really appreciate your help (no, really; after
tormenting about this on my own for some time now it really means a
lot). I'll compute RV coefficients and R-squared values following all
your suggestions!
many thanks,
Manabu
Manabu Sakamoto, PhD
Postdoctoral Research Associate
School of Earth Sciences
University of Bristol
Bristol, UK, BS8 1RJ
Tel: +44 (0) 117 954 5421
Fax: +44 (0)117 925 3385
Email: m.sakam...@bristol.ac.uk
On 29 Jun 2011, at 17:31, morphmet wrote:
-------- Original Message --------
Subject: Re: correlation coefficient between two or more matrices
Date: Wed, 29 Jun 2011 03:36:33 -0400
From: andrea cardini <alcard...@interfree.it>
To: morphmet@morphometrics.org
Dear Manabu,
I agree with you that a regression might be more appropriate if you have
predictors with which you want to predict a set of dependent variables.
Computing the variance predicted by the whole set of independent variables
is straightforward and part of the output of programs like MorphoJ and
TPSRegr.
It's easy to do also manually:
1) save the predictions of the regression (unstandardized);
2) compute their variances one at a time and sum them up;
3) compute the variances of each of the dependent variables and sum them up;
4) divide 2) by 3) and multiply by 100 to get the % of variance explained.
PAST does not do, if I am correct, multivariate (many dependent variables)
multiple (many predictors) regressions. R certainly does it as well as
commercial software like SPSS, NTSYS and many others.
Good luck.
Cheers
Andrea
At 12:28 28/06/2011 -0400, you wrote:
-------- Original Message --------
Subject: correlation coefficient between two or more matrices
Date: Fri, 24 Jun 2011 04:58:54 -0400
From: Manabu Sakamoto <m.sakam...@bristol.ac.uk>
To: morphmet@morphometrics.org
Dear list,
I am trying to get the correlation coefficient (or coefficient of
determination) between two or more matrices. I have one matrix of
response variables and one or more matrices of predictor variables. My
goal is to compute the proportion of variance in my response variable
that can be explained by the predictor variable/variables. I think a
regression is appropriate for this situation rather than a 2B-PLS.
However, I am at a loss as to how one should do this. The software
implementations that I am aware of (PAST, R) seems to compute the
correlation coefficient and R-squared values for each response variable
at a time, but I want a single R-squared value for the correlation
between the matrices.
I know the Mantel test in PAST computes the Pearson's correlation
coefficient but I've been reading that the Mantel test has low power and
should be avoided.
Can someone on the list give me suggestions?
many thanks in advance,
best wishes,
Manabu
Manabu Sakamoto, PhD
Postdoctoral Research Associate
School of Earth Sciences
University of Bristol
Bristol, UK, BS8 1RJ
Tel: +44 (0) 117 954 5421
Fax: +44 (0)117 925 3385
Email: m.sakam...@bristol.ac.uk
Dr. Andrea Cardini
Researcher in Animal Biology
Dipartimento di Biologia, Universitá di Modena e Reggio Emilia, via Campi
213, 41100, Modena, Italy
tel: 0039 059 2055017 ; fax: 0039 059 2055548
Honorary Fellow
Functional Morphology and Evolution Unit, Hull York Medical School
University of Hull, Cottingham Road, Hull, HU6 7RX, UK
University of York, Heslington, York YO10 5DD, UK
Adjunct Associate Professor
Centre for Forensic Science , The University of Western Australia
35 Stirling Highway, Crawley WA 6009, Australia
E-mail address: alcard...@interfree.it, andrea.card...@unimore.it,
andrea.card...@hyms.ac.uk
Webpage: http://sites.google.com/site/hymsfme/drandreacardini
Datasets:
http://ads.ahds.ac.uk/catalogue/archive/cerco_lt_2007/overview.cfm#metadata
