Coefficients of log of parts in a balance matrix,
(+1) for numerator and (-1) for denominator, are transformed
into the corresponding isometric log-ratio (ilr) coefficients
Usage
B2i(bal, balnames=FALSE)
Value
a \(D \times (D-1)\) matrix giving the coefficients of the ilr transforms
Arguments
bal
a (\(D-1\) x \(D\)) balance matrix with cells +1, 0 or -1.
balnames
logical, if TRUE, balance names are attributed to ilr transforms; if FALSE (default) ilr transforms are numbered \(ilr1\) to \(ilrD1\), where \(D1=D-1\) and \(D\) is the number of parts.
Details
Two scalars multiplying positive and negative cells respectively are defined for each row of the matrix \(bal\) in such a way that the resulting matrix defines the ilr transformation to apply to the log of a compositional vector. The output transformation matrix is transposed for application to a compositional dataset where the compositions are the rows.
References
Pawlowsky-Glahn, V., J. J. Egozcue, and R. Tolosana-Delgado (2007). Lecture Notes on Compositional Data Analysis.
bal <- matrix(c(1,-1,0,1,1,-1),nrow=2, byrow=TRUE)
colnames(bal) <- paste("l.P",1:3,sep="")
bal
B2i(bal)
rownames(bal) <- paste("B",1:2,sep="")
bal
B2i(bal,balnames=TRUE)
B2i(bal)