Comparing matrices without a Mantel Test

#1
Hello everyone,

I am a biologist at a University and have stumbled across an interesting problem. My data is organized into matrices. The matrices are rectangular not triangular. I would like to compare matrices to each other but can't find a statistical way to do it that is not the Mantel test. Mantel tests only work when measuring distance and on triangular matrices. I am working with neither.

If anyone knows a way to solve this problem that would be great! I will answer any other questions about my data you may have.
 
#3
I am currently reading a paper on this method. It still seems to apply to only distance matrices. Is this true? My matrices aren't distance matrices.
 

bugman

Super Moderator
#5
If you have access to the PRIMER software, you can do this will the function RELATE.

You say you want to compare matrices. A bit more info regarding the data would be insightful and some specific idea about what you are trying to do would also be good.

Are you trying to look at community differnces between sites, times?
What are the two matrices? - before / after data. Different collection / enumeration methods....?
 
#6
I study birds and am using this matrices to look at their singing behavior. Both males and females sing in my study species and they combine their songs together. The matrices have male song types along the y-axis of the matrix and female song types along the top of the matrix. The numbers in the cells of the matrix indicate how often certain combinations of the male and female song types have occurred. These matrices also contain a lot of cells with zeros in them. I am trying to compare the same pairs singing behavior between different years.

I have many of these matrices and want to be able to compare them in an efficient manner.
 

bugman

Super Moderator
#7
I strongly recomend looking into PRIMER or PERMANOVA for this.

PERMANOVA is available as an add-on to PRIMER V6. If you dont have PRIMER, the functions (most of them I think) are available in the R-packages: MASS and VEGAN. Permanova can also be implemented in R (adonis).

These tests will allow you to combine all of you matrices as one - adding in "years" "gender" etc... as factors, thus removing the need to have several matrices.

If you are familiar with R, I can nudge you in the right direction. If you are at a university, I would be surprised if there wasn't a copy of PRIMER lurking around.

I'll post an introduction PERMANOVA next...
 

bugman

Super Moderator
#8
... here it is.

There is a link to Marti Andersons home page and there you will find a free DOS version to accompany the manual.:tup: