VCA (version 1.3.4)

Csweep: Calling C-implementation of the SWEEP-Operator

Description

Function calls a fast C-implementation of the SWEEP operator using the transpose of the original augmented matrix \(X'X\) (see getSSQsweep).

Usage

Csweep(M, asgn, thresh = 1e-12, tol = 1e-12, Ncpu = 1)

Arguments

M

(matrix) matrix, representing the augmented matrix \(X'X\)

asgn

(integer) vector, identifying columns in \(M\) corresponding to variables, respectively, to their coefficients

thresh

(numeric) value used to check whether the influence of the a coefficient to reducing the error sum of squares is small enough to conclude that the corresponding column in \(X'X\) is a linear combination of preceding columns

tol

(numeric) value used to check numerical equivalence to zero

Ncpu

(integer) number of cores to be used for parallel processing (not yet used)

Value

(list) with two elements:

SSQ

(numeric) vector of ANOVA sum of squares

LC

(integer) vector indicating linear dependence of each column

Details

Transposing prior to applying the SWEEP-operator speeds up things since the complete matrix is stored in memory in consecutive manner.

This is an utility-function not intended to be called directly.

References

Goodnight, J.H. (1979), A Tutorial on the SWEEP Operator, The American Statistician, 33:3, 149-158

Examples

# NOT RUN {
# use example from 'lm' Rdoc
ctl    <- c(4.17,5.58,5.18,6.11,4.50,4.61,5.17,4.53,5.33,5.14)
trt    <- c(4.81,4.17,4.41,3.59,5.87,3.83,6.03,4.89,4.32,4.69)
group  <- gl(2, 10, 20, labels = c("Ctl","Trt"))
weight <- c(ctl, trt)
lm.D9  <- lm(weight ~ group)
anova(lm.D9)

# create augmented matrix
X  <- model.matrix(lm.D9)
Xt <- t(X)
y  <- matrix(weight, ncol=1)
yt <- t(y)
M  <- rbind(	cbind(as.matrix(Xt%*%X), as.matrix(Xt%*%y)), 
		        cbind(as.matrix(yt%*%X), as.matrix(yt%*%y)))
swept <- VCA:::Csweep(M, asgn=c(0,1))					
LC    <- swept$LC
SS    <- swept$SSQ
SSQ   <- abs(diff(SS))
SSQ   <- c(SSQ, tail(SS,1))
SSQ		# compare to column "Sum Sq" in 'anova(lmD9)' output
# }