harmonize: Harmonization by calibration

In StratifiedSampling: Different Methods for Stratified Sampling

Harmonization by calibration

Description

This function harmonize the two weight schemes such that the totals are equal.

Usage

harmonize(X1, d1, id1, X2, d2, id2, totals)

Arguments

X1	A matrix, the matching variables of sample 1.
d1	A numeric vector that contains the initial weights of the sample 1.
id1	A character or numeric vector that contains the labels of the units in sample 1.
X2	A matrix, the matching variables of sample 2.
d2	A numeric vector that contains the initial weights of the sample 1.
id2	A character or numeric vector that contains the labels of the units in sample 2.
totals	An optional numeric vector that contains the totals of the matching variables.

Details

All details of the method can be seen in the manuscript: Raphaël Jauslin and Yves Tillé (2021) <arXiv:>.

Value

A list of two vectors, the new weights of sample 1 (respectively new weights of sample 2).

Examples

#--- SET UP

N = 1000
p = 5
X = array(rnorm(N*p),c(N,p))

n1=100
n2=200

s1 = sampling::srswor(n1,N)
s2 = sampling::srswor(n2,N)

id1=(1:N)[s1==1]
id2=(1:N)[s2==1]

d1=rep(N/n1,n1)
d2=rep(N/n2,n2)

X1 = X[s1==1,]
X2 = X[s2==1,]

re <- harmonize(X1,d1,id1,X2,d2,id2)

colSums(re$w1*X1)
colSums(re$w2*X2)

#---  if the true totals is known

totals <- c(N,colSums(X))
re <- harmonize(X1,d1,id1,X2,d2,id2,totals)

colSums(re$w1*X1)
colSums(re$w2*X2)
colSums(X)

StratifiedSampling documentation built on Oct. 26, 2022, 5:09 p.m.