-------- Original Message --------
Subject: controlling for tooth wear
Date: Wed, 7 Dec 2011 16:16:27 -0500
From: Rodrigo Lima <rodrigo.l...@mail.mcgill.ca>
To: morphmet@morphometrics.org <morphmet@morphometrics.org>
Hello morphometricians,
I would be really thankful if you could help me with the following problem:
I'm studying differences in shape and size of teeth between different
populations of voles, and trying to explain this variation through
environmental and geographical variables. Tooth wear is a great source
of variation, so I'm trying to get rid of the effect of wear. I tried
two very similar approaches:
1 - First I classified all teeth into 5 stages of tooth wear (0 being
the "brand new" teeth, 4 being the "very worn-out" teeth). Then I
performed a linear regression using scores on principal components as
the dependent variables and tooth wear stage as the independent
variable. I took the residuals and used them as the new variables
against which I regressed my environmental and geographical variables.
2 - I did the same thing as in 1, but instead of residuals from a linear
regression I used the residuals from a GLM.
3 - I did the same thing as in 1 and 2 with the size variable (occlusal
surface area).
The problem: the results of the regressions of residuals on
environmental and geographical variables have very low r2 (adjusted r2
always around 0.04). If I ignore tooth wear I get a higher r2 (around
0.25). So: does this mean that practically all the variation in my
sample is due to tooth wear? Am I doing something wrong here?
I thank in advance all of those who kindly use their time to read this post.
Best wishes,
Rodrigo
MSc student
McGill University
(rodrigo.l...@mail.mcgill.ca)
