Seemingly different results on FA from MorphoJ and geomorph, where am I wrong

276 views
Skip to first unread message

lv xiao

unread,
Dec 20, 2017, 10:10:37 AM12/20/17
to geomorph R package
Hi, I am using both MorphoJ and geomorph to calculate FA for the same set of data, but it seems that I get different results.

The data is as attached. There are 82 3D specimens with object symmetry, each digitized twice.

In MorphoJ, following Procrustes ANOVA for object symmetry, the following results are obtained:
Classifiers used for the Procrustes ANOVA:
Individuals: Individual


Centroid size:
Effect                 SS            MS        df        F      P (param.)
Individual   88301.909308     1090.147028      81    4457.88      <.0001
Residual        20.052584        0.244544      82


Shape, Procrustes ANOVA:
Effect              SS              MS         df        F      P (param.)
Individual     0.41479207    0.0000701492     5913      11.29      <.0001
Side           0.00249669    0.0000390107       64       6.28      <.0001
Ind * Side     0.03220384    0.0000062122     5184       3.03      <.0001
Residual       0.02306261    0.0000020529    11234



Shape, MANOVA tests of effects:


Symmetric component of shape variation:
Effect        Pillai tr.   P (param.)
Individual      59.58       <.0001
Note: the test for 'Individual' used the symmetric component of the residual as the 'error' effect.


Asymmetry component of shape variation:
Effect        Pillai tr.   P (param.)
Side             0.94       0.0003
Ind * Side      43.35       <.0001

Whereas in geomorph, I executed the following line of code:
summary(bilat.symmetry(A = shape, ind = ind, object.sym = TRUE, land.pairs=pair, data = gdf, RRPP = TRUE, iter = 19, print.progress = T))
and the output is as follows:
Call:
bilat
.symmetry(A = shape, ind = ind, object.sym = TRUE, land.pairs = pair,  
    data
= gdf, iter = 19, RRPP = TRUE, print.progress = T)




Symmetry (data) type: Object


Type I (Sequential) Sums of Squares and Cross-products
Randomized Residual Permutation Procedure Used
20 Permutations


Shape ANOVA
         
Df      SS        MS     Rsq       F      Z   Pr(>F)  
ind      
81 0.41755 0.0051550 0.87785 12.8907 3.7663   0.05 .
side      
1 0.00250 0.0025036 0.00526  6.2605 4.1005   0.05 .
ind
:side 81 0.03239 0.0003999 0.06810                        
---
Signif. codes:  0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 1

I understand that different algorithms were used in MorphoJ and geomorph, but I do not understand why even the degrees of freedom are different.

I am wondering if my code in geomorph is somewhere wrong. Please help me out for this issue. Thank you very much!

Example 13_object_ symmetry_sheep_example.txt

Adams, Dean [EEOBS]

unread,
Dec 20, 2017, 10:16:35 AM12/20/17
to geomorph-...@googlegroups.com

A few quick points.

 

1: look at the SS and MS for the shape effects. They are identical in both up to rounding error. Thus, any differences are downstream of the algebra and rest in how significance is assessed.

 

2: In that regard, MorphoJ uses parametric significance testing based on MANOVA. These are laden with many untenable assumptions for shape data: multivariate normality, homoscedasticity, assumptions linking multivariate to univariate test measures, etc.  By contrast, geomorph uses permutation tests to evaluate significance.

 

3: You are only using 19 permutations for significance testing in geomorph: that makes no sense. Bump that up to 999 or 9999 for more precise p-values.


Dean

 

Dr. Dean C. Adams

Professor

Department of Ecology, Evolution, and Organismal Biology

       Department of Statistics

Iowa State University

www.public.iastate.edu/~dcadams/

phone: 515-294-3834

--
You received this message because you are subscribed to the Google Groups "geomorph R package" group.
To unsubscribe from this group and stop receiving emails from it, send an email to geomorph-r-pack...@googlegroups.com.
To post to this group, send email to geomorph-...@googlegroups.com.
Visit this group at https://groups.google.com/group/geomorph-r-package.
To view this discussion on the web, visit https://groups.google.com/d/msgid/geomorph-r-package/85b4cb6b-08be-4e14-9953-e74fc9689098%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

lv xiao

unread,
Dec 21, 2017, 4:48:32 AM12/21/17
to geomorph R package
Thank you so much Prof. Dean, I now get it.

To unsubscribe from this group and stop receiving emails from it, send an email to geomorph-r-package+unsub...@googlegroups.com.
To post to this group, send email to geomorph...@googlegroups.com.

Reply all
Reply to author
Forward
0 new messages