-------- Original Message --------
Subject: Re: Variance explained by size
Date: Thu, 1 Sep 2011 13:00:00 -0400
From: Joseph Kunkel <[email protected]>
To: [email protected]
Wilks Lambda is a distribution function for analysis of dispersion which
is a generalization of analysis of variance where you have multiple Y
variables in the General Linear Model Y = XB.
This is the situation you have in landmark analysis with multiple (x,y)
or (x,y,z) landmarks for each specimen. Wilks Lambda can be transformed
into a Chi-Square distribution or an F-distribution so that those
distribution tables of significance can be used.
See C. Radhakrishna Rao (1965) Linear Statistical Inference and Its
Applications, 1st Edition, for a good discussion.
Joe Kunkel
On Sep 1, 2011, at 12:40 PM, morphmet wrote:
-------- Original Message --------
Subject: Re: Variance explained by size
Date: Thu, 1 Sep 2011 12:09:41 -0400
From: Francisco prevosti <[email protected]>
Reply-To: Francisco prevosti <[email protected]>
To: [email protected] <[email protected]>
I have a question in relationship to this post.
TpsReg report the Wilks lambda for the multivaritate regression? it is
possible to calculate the explained variation (i.e. 1-wilks lambda), as
is doing in other contexts??? or what is the meaning of wilks lambda in
the multivariate regression context???
sincerely,
pancho
Francisco J. Prevosti
División Mastozoología
Museo Argentino de Ciencias Naturales "Bernardino Rivadavia" - CONICET
Av. Angel Gallardo 470 - C1405DJR -
Buenos Aires - Argentina -
Tel/Fax.: (5411) 4982-0306 / 1154 / 5243 / 4494 - Int. 210
http://www.macn.secyt.gov.ar/
https://sites.google.com/site/southamericancarnivores/home
------------------------------------------------------------------------
*From:* morphmet <[email protected]>
*To:* morphmet <[email protected]>
*Sent:* Thursday, September 1, 2011 11:40 AM
*Subject:* Re: Variance explained by size
-------- Original Message --------
Subject: Re: Variance explained by size
Date: Thu, 1 Sep 2011 10:33:58 -0400
From: andrea cardini <[email protected]
<mailto:[email protected]>>
To: [email protected] <mailto:[email protected]>
Dear Brian,
if what you're interested to know is the % of variance explained by size
(your independent predictor variable), you just need to:
1) save the predictions of the multivariate regression (MorphoJ, NTSYSpc,
SPSS, SAS, R etc. they all allow you to do so);
2) do their variances and sum them up;
3) do the variances of the original shape variables (e.g., all the PCs of
the shape coordinates) and sum them up;
4) do the ratio between 2) and 3) (and multiply it by 100): that will give
you the same % of explained variance you find in MorphoJ, Morphologika or
the one you get from TPSRegr by doing 100 minus the unexplained variance
(in the report window of TPSRegr).
There are other ways to do these computations but this is very easy and you
just need an xls spreadsheet for doing the variances, sums and ratio. With
Gnumeric (freeware), you can directly drag the txt file exported from
MorphoJ into the spreadsheet and you'll have the variables ready for the
computations.
There might be exceptions but I would generally avoid to regress one PC at
a time onto size or any other predictor. Shape is inherently multivariate
and PCs are derived only to maximize total sample variance (and not to,
say, best fit a size predictor).
Be careful if you try to get the explained % from a standard statistical
software not to confuse a multiple regression (many predictors and only one
predicted variable - e.g., you want to predict size based on temperature,
humidity, rainfall etc.) with a multivariate one (just one predictor and
many predicted variables - e.g., allometry when you regress shape variables
all together onto size). The Rsq one generally finds in the standard
statistical software is the one for the multiple regression (i.e., not the
one you may want for testing allometry).
Good luck.
Cheers
Andrea
At 10:31 29/08/2011 -0400, you wrote:
>
>
>-------- Original Message --------
>Subject: Variance explained by size
>Date: Fri, 26 Aug 2011 17:29:21 -0400
>From: brian boivin <[email protected] <mailto:[email protected]>>
>To: [email protected] <mailto:[email protected]>
>
>Hi,
>In Geometrics morphometrics for Biologist : A Primer (pg7) it says:"In
>the two species mentionned above (in which PC1 accounts for 99.4% of
>the variance), SIZE explains 71% of the variance in SHAPE in one
>species, but only 21.7% in the other."
>I did not find any information in the book to explains the impact of
>size on the variance in shape. Did I miss something?
>How can one calculate the % of variance in shape explained by size?
>Email:[email protected] <mailto:[email protected]>
>Thank you for your time
>B.B
>
>
>
>
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: [email protected] <mailto:[email protected]>,
[email protected] <mailto:[email protected]>,
[email protected] <mailto:[email protected]>
Webpage: http://sites.google.com/site/hymsfme/drandreacardini
Datasets:
http://ads.ahds.ac.uk/catalogue/archive/cerco_lt_2007/overview.cfm#metadata
-·. .· ·. .><((((º>·. .· ·. .><((((º>·. .· ·. .><((((º> .··.· >=-
=º}}}}}><
Joseph G. Kunkel, Professor
Biology Department
University of Massachusetts Amherst
Amherst MA 01003
http://www.bio.umass.edu/biology/kunkel/
