-------- Original Message --------
Subject: Re: Sexual dimorphism: Procrustes ANOVA vs DFA
Date: Thu, 4 Aug 2011 05:12:34 -0400
From: Philipp Mitteröcker <[email protected]>
To: [email protected]
Dear Louis,
- Discriminant function analysis per se does not give you a p-value. I
guess you performed a Hotelling's T-square test for each strain
separately. 16 specimens per strain is quite a small sample size. As a
p-value depends to a large amount on sample size, your results thus may
not be significant.
- If you use ANOVA for several strains, you utilize a larger sample than
if you analyze each strain separately. Thus it is possible that for each
strain sexual dimorphism is not significant, but taken all strains
together it is significant. However, this two-way ANOVA approach assumes
that sexual dimorphism is the same in all strains.
- If sexual dimorphism is similar in all strains, you may get a
significant main effect for sex, but no significant interaction term sex
x strain. If sexual dimorphism varies considerably across strains, the
interaction may be significant but not the main effect.
Hope this helps,
Philipp
Am 03.08.2011 um 18:14 schrieb morphmet:
-------- Original Message --------
Subject: Sexual dimorphism: Procrustes ANOVA vs DFA
Date: Wed, 3 Aug 2011 05:38:23 -0400
From: Louis Boell <[email protected]>
To: <[email protected]>
Dear all,
I would like to ask for advice with the following problem:
in a study of mouse mandible shape using Procrustes coordinates, looking
at differences between inbred strains, I want to test for sexual
dimorphism (with sample sizes of about 8 males and 8 females each for
each strain and 14 landmarks).
I did two things:
1)to see whether there are shape differences between males and females
within strains, I used discriminant function analysis between females
and males for each strain separately. I found no significant differences
at all.
2)I performed Procrustes ANOVA on pairs of strains with sex and strain
as main effects. Depending on the combination of strains, I usually get
a significant strain effect, and either no significant sex effect, but a
significant strain x sex interaction, or a significant sex effect, but
no significant interaction.
My problem consists of my limited understanding of ANOVA and hence
limited ability to interpret the results. I have consulted the
literature, but textbooks are either cryptic to me or too superficial,
and they do not cover the answers to my questions:
1)Most often ANOVA is used in this context, and reviewers want me to use
ANOVA. Why is ANOVA “better” than DFA in this context?
2)Why do I get significant sex effects or sex x strain interactions with
ANOVA, while I find no significant sexual dimorphisms with DFA?
3)Why do I get, in almost all cases, either a significant sex effect, or
a significant sex x strain interaction, but almost never both sex effect
and interaction significant, and what does each result tell me about my
data?
I would greatly appreciate your help.
Best wishes,
Louis Boell
Louis Boell
Louis Boell
MPI für Evolutionsbiologie, Plön
August-Thienemannstr.2
D-24306 Plön
Tel.: 0049 4522 763 280
___________________________________
Dr. Philipp Mitteroecker
Department of Theoretical Biology
University of Vienna
Althanstrasse 14
A-1090 Vienna, Austria
Tel: +43 1 4277 56705
Fax: +43 1 4277 9544
email: [email protected]
homepage: http://theoretical.univie.ac.at/people/mitteroecker
