The mvpart Package - NexTag Supports Open Source Initiatives
The mvpart Package - NexTag Supports Open Source Initiatives
The mvpart Package - NexTag Supports Open Source Initiatives
You also want an ePaper? Increase the reach of your titles
YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.
6 gdist<br />
gdist<br />
Dissimilarity Measures<br />
Description<br />
Usage<br />
<strong>The</strong> function computes useful dissimilarity indices which are known to have a good rank-order relation<br />
with gradient separation and are thus efficient in community ordination with multidimensional<br />
scaling.<br />
gdist(x, method="bray", keepdiag=FALSE, full=FALSE, sq=FALSE)<br />
Arguments<br />
x<br />
method<br />
keepdiag<br />
full<br />
sq<br />
Data matrix<br />
Dissimilarity index<br />
Compute amd keep diagonals<br />
Return the square dissimilarity matrix<br />
Square the dissimilarities – useful for distance-based partitioning<br />
Details<br />
<strong>The</strong> function knows the following dissimilarity indices:<br />
euclidean d jk = √∑ i (x ij − x ik ) 2<br />
manhattan d jk = ∑ i |x ij − x ik |<br />
gower d jk = ∑ |x ij−x ik |<br />
i max i − min i<br />
canberra d jk = 1 ∑ |x ij−x ik |<br />
N−Z ∑ i x ij+x ik<br />
bray d jk =<br />
|xij−x ik|<br />
∑i (xij+x ik)<br />
i ∑<br />
kulczynski d jk = 1 − 0.5(<br />
i min(xij,x ik)<br />
∑i xij +<br />
∑<br />
∑<br />
i min(xij,x ik)<br />
maximum d jk = max i |x ij − x ik |<br />
binary d jk = ∑ i |x ij > 0 − x ik > 0|<br />
chord d jk = √∑ i (x ij − x ik ) 2 / ∑ i (x ij + x ik ) 2<br />
i x ik<br />
)<br />
where N − Z is the number of non-zero entries.<br />
Infamous ”double zeros” are removed in Canberra dissimilarity.<br />
Euclidean and Manhattan dissimilarities are not good in gradient separation without proper standardization<br />
but are still included for comparison and special needs.<br />
Some of indices become identical or rank-order similar after some standardizations.