Skip to content

help with analysis

3 messages · Kátia Emidio, Pedro Higuchi, Dave Roberts

Original
11 days later
Dave Roberts
# 
Hi K?tia,

    Sorry to be so late; just back in the office.

    Pedro is correct that adonis will help establish statistical 
significance to potential differences, but I think NMDS could still be 
very helpful.  One approach would be to code the sites by glyph (e.g. 
site 1 = circle, site 2 = triangle, etc.) and then to draw arrows from 
the first date to the second, second to third, etc, for each site.

    In labdsv you could do this if your nmds is called nmds.object

 > plot(nmds.object,type=n')

to draw the axes but not the points, and then use

 > points(nmds.object,site==1,pch=1)
 > points(nmds.object,site==2,pch=2)

etc.  Then, depending on how the sites are sorted, the arrows could be drawn

 > 
arrows(nmds.object$points[1,1],nmds.object$points[1,2],nmds.object$points[2,1],
       nmds.object$points[2,2])

to draw an arrow form the first point to the second.  YOu might need to 
mess with the parameters of arrows() to get the arrow sizes you want.

    In vegan you could do

 > nmds.plot <- plot(metaMDS(dissimilarity or taxon matrix),type='n')
 > points(nmds.plot$sites[site==1])
 > points(nmds.plot$sites[site==2],pch=2)
etc
 > arrows(nmds.plot$sites[1,1],nmds.plot$sites[1,2],
    nmds.plot$sites[2,1],nmds.plot$sites[2,2])

etc.

Hope that helps, Dave