-------- Original Message --------
Subject: Re: predictive shape variation
Date: Wed, 29 Jun 2011 06:15:22 -0400
From: Ondřej Mikula <[email protected]>
To: [email protected]
Thank you, I have known that the task is non-standard, but perhaps it
is even ill-defined. I just felt it does matter whether equally good
prediction would be achieved with a minor part of shape variation.
2B-PLS would be the natural choice but I have used Procrustes
residuals to reproduce a matrix of genetic distances in a kind of
multidimensional scaling. It was not a classical regression analysis
with two sets of variables.
Ondrej Mikula
On 28 June 2011 21:54, morphmet <[email protected]>
wrote:
-------- Original Message --------
Subject: RE: predictive shape variation
Date: Tue, 28 Jun 2011 14:56:53 -0400
From: F. James Rohlf <[email protected]>
Reply-To: <[email protected]>
To: <[email protected]>
One normally estimates the proportion of the variance of the dependent
variables that is accounted for by regression on a set of one or more
independent variables (covariates). I am not sure if it makes sense to do
it the other way around - to see how much of the variation in the
independent variables is useful for predicting the dependent variables. One
could look at the problem in a more symmetric way by using two-block partial
least squares and seeing how much variation is in common.
-------------------
F. James Rohlf, John S. Toll Professor
Dept. Ecology & Evolution, Stony Brook University
Please consider the environment before printing this email
-----Original Message-----
From: morphmet [mailto:[email protected]]
Sent: Tuesday, June 28, 2011 2:30 PM
To: morphmet
Subject: predictive shape variation
-------- Original Message --------
Subject: predictive shape variation
Date: Tue, 28 Jun 2011 13:27:16 -0400
From: Ondřej Mikula <[email protected]>
To: [email protected]
Dear morphometricians,
I have an analysis where shapes are used as covariates. With regression
coefficients in hand I wish to quantify what proportion of shape variation
was
actually predictive. Could you give me a hint how to proceed?
Best regards
Ondrej Mikula
Institute of Animal Physiology and Genetics Brno, Czech Republic