Cost function of 2 alternative classifications of rows in vegetation data

Description

Given 2 alternative classifications (g groups) of rows in a data frame of vegetation data, confusion matrix, C, is derived first. Using the first classification a matrix of row centroids is derived (using function centroid) of wich a g by g distance matrix, W, is computed (correlation transformed to distance). Cost factor, cf, is the sum of element by element multiplication of C and W respectively, cf=sum(CW).

Usage

ccost(veg, oldgr, newgr, y,...)
ccost2(veg,oldgr, newgr, y)

## Default S3 method:
ccost(veg, oldgr, newgr, y,...)
## S3 method for class 'ccost'
print(x,...)

Arguments

Details

Cost factor cf has range 0 (both classifications identical) to n (number of rows), where n is the worst case of misclassification.

Value

An output list of class "ccost" with at least the following intems:

Author(s)

Otto Wildi

References

Ripley, B. D. 1996. Pattern recognition and neural networks. Cambridge: Cambridge University Press.

Venables, W. N. & Ripley, B. D. 2010. Modern applied statistics with S. Fourth Edition. Springer, NY.

Wildi, O. 2013. Data Analysis in Vegetation Ecology. 2nd ed. Wiley-Blackwell, Chichester.

Examples

# First, groups of releves are formed by cluster analysis
require(vegan)
dr<- vegdist(nveg^0.5,method="bray")      # dr is distance matrix of rows
o.clr<- hclust(dr,method="ward")          # this is clustering
oldgr<- cutree(o.clr,k=3)                 # 3 row groups formed
oldgr                                     # this displays initial classification:
# 2  4  6  9 10 18 25 27 39 49 50
# 1  2  1  3  2  3  1  2  3  1  3

# For simplicity we assume that row "2" and "50" change memebership:
newgr<- c(2,2,1,3,2,3,1,2,3,1,1)
o.ccost<- ccost(nveg,oldgr,newgr,y=0.5)           # does square root transformation
# Default method releasing cf
o.ccost                                           # displays C and W (see above)

Results