-------- Original Message --------
Subject: Re: multv. regression for allometry
Date: Wed, 20 Oct 2010 11:38:48 -0400
From: thimacek <[email protected]>
To: [email protected]
I can't really help with the results in MorphoJ and SPSS differing, but
as for your concern regarding the appropriateness of carrying out a
multivariate regression analysis with only one independent variable, I'd
say you can rest assured that it is not a problem.
Keeping capitals as matrixes and lower case letters as vectors: what you
actually have is not
Y = a + b.x, but rather
Y = A + B.x
The general model for linear multivariate regression (following the most
used terminology) is one in which the dependent variables (the matrix Y
above) are modelled/ defined as a function of a single variable 'x';
instead of the usual intercept (a) and angular coefficient (b) from
simple linear regression, one has a matrix of intercepts (A) and a
matrix of coefficients (B).
Multiple regression is the name usually used for when one's modelling
the effects of a number of independent variables on one (or more)
dependent variables , e.g. y = a + b.x + c.z + d.w or, if the effect is
being analysed on multiple dependent variables: Y = A + B.x + C.z + D.w
(...)
That... may have been a bit confusing, but should be the gist of it.
Can someone shed further light on the point (or correct me if I'm
mistaken, of course)?
------------------------------------------------------------------------
Em 20/10/2010 09:23, *morphmet < [email protected] >*
escreveu:
-------- Original Message --------
Subject: multv. regression for allometry
Date: Wed, 20 Oct 2010 05:23:04 -0400
From: Deniz SIGIRLI
To: [email protected]
Dear group members,
In allometry studies there is one independent variable (x:centroid size)
and more than one dependent variables (Y is a matrix for shape
parametres) and the model is as Y=a+bx. But as I know, one of the
assumptions of multivariate regression analysis is that, both the
dependent and independent variables must be multiple (Y=A+BX ) .
And the results of multivariate regression analysis in MorphoJ program
and SPSS are different (for tangent coord. and centroid size).
I dont understand this situation. What are your opinions about this topic.
