Skip to content

your suggestions in MRMs

4 messages · Kristi Glover, Sarah Goslee

Kristi Glover
# 
Hi R Users,
I was trying to  perfom multiple regression on resemblance matrices (MRMs). This technique in avaiable  in "ecodist" package  and looked at the example data to know how I need to organize my data set. I think the data is distance matrix but I was wondering the rows name. For example, there are (these are the subset of the data of "graze") 
         sitelocation forestpct
1.1.2001    12.187743     63.88
1.2.2001    12.186077     71.33
2.1.2001    12.406362     72.45
2.2.2001    12.416265     77.13
3.1.1998     8.409213     18.35

if we look at the first row, 1.1.2001: sitelocation (column) is 12.187.. which is the euclidean distance between two points (XY cordinates). But I was confused at the row name where 1.1.2001 which is to me is site1 and site1 of 2001. Isn't it supposed to be "0" if both are the same site.  I think I misunderstood it. Any one can help me about what it is? 

I put the example for your reference

install.packages("ecodist")
library(ecodist)
data(graze)
graze[1:5,1:2]

Thanks for your help

KG
Sarah Goslee
# 
Kristi,

The row names are utterly arbitrary. Each row is a separate site, and
sitelocation is a location variable (both intended to conceal the
absolute location, which is confidential since it's on private
property). It is NOT the Euclidean distance, nor is a row representing
a pair of sites.

If you look at the full example, dist() is used to calculate the
Euclidean distance as part of the MRM code.

Sarah
On Wed, Apr 8, 2015 at 1:49 PM, Kristi Glover <kristi.glover at hotmail.com> wrote:

  
    

      
      
    

  


  


      

    

    

      
      

        


  

        

      Kristi Glover
    

    

      
      #
    

  


  

      
Hi Sarah,

Thanks for the reply. I again looked at the example, but I did not find the way to
calculate the location variable. All example in the document
("ecodist") has the Euclidean distance. did not find the example to
conceal the absolute location. There are some example:

page 21 and 22 : iris.md <- distance(iris[,1:4], "mahal"): it is
calculated distance matrix using four variables (I understood this one)

page 26: space.d <- distance(space, "eucl"), here space is XY coordinate,
so on



I am just wondering how I can conceal my absolute location (XY) and make my
data compatible to your data set so that I can use your functions. I am sorry for bothering you, Sarah. 

example of my data set: column names (site,
XY coordinates, temp, years)

site1, 25.01;34.78 (XY cordicnate),
35degree celcius; year2001

site1, 25.01;34.78 (XY cordicnate), 33degree celcius;
year2002

site2, 25.05; 35.56 (XY coordinate); 37degree
celcius; year2001

site2, 25.05; 35.56 (XY coordinate);
32degree celcius; year2002





Thanks

 

=======

            

      
      
      
    

        

  
  


  


      

    

    

      
      

        


  

        

      Sarah Goslee
    

    

      
      #
    

  


  

      
Kristi,

You're completely missing the point, I think.

Instead of providing X,Y coordinates in the sample dataset graze
within the ecodist package, I provided one location, X if you'd like,
called sitelocation.

If you look at the example in ?MRM,

data(graze)
LOAR10.mrm <- MRM(dist(LOAR10) ~ dist(sitelocation) + dist(forestpct),
data=graze, nperm=100)

This is a toy example,with only one species, geographic distance, and
another potential explanatory variable.

dist(LOAR10) - species distance; you'd use whatever set of species
you're studying

dist(sitelocation) - geographic distance, you'd use your x and y
coordinates as in dist(xy.matrix)

dist(forestpct) - you'd use whatever variable or variables are
appropriate for your study


I concealed the location of these sites BEFORE I made the data public,
which is why there's one location variable instead of x,y coordinates.

Each row of graze is a single site, with some information associated.

You don't need to conceal your location for your own analysis. You
just need to use the appropriate data within the dist() command to
calculate the distances, and with the appropriate distance metrics if
Euclidean isn't appropriate for your data.

Reading the papers cited in ?MRM might help you understand the logic a
bit better.

Sarah

        
On Wed, Apr 8, 2015 at 3:01 PM, Kristi Glover <kristi.glover at hotmail.com> wrote: