# saEqualClus: Optimal Equal Sized Strata Formation and Allocation Using... In jlisic/saAlloc: Optimal Allocation of Sampling Units to Form Strata using Simulated Annealing

## Description

This function minimizes the sample size for equally sized strata by exchanging sampling units between strata. Optimization is performed under variance constraints for known administrative variables.

## Usage

 1 2 3 4 5 6 7 8 9 10 11 saEqualClus( x, label, iterations=1000, cooling=0, segments=rep(1,nrow(x)), controlVariable=rep(1,nrow(x)), targetVar, varWithin=rep(0,ncol(x)), tolSize=0.05 )

## Arguments

 x (Required) A matrix where each row is a sampling unit and each column is a characteristic of the sampling unit that is to be reduced in optimal sampling size. label (Required) Initial stratum assignment. iterations Number of iterations to perform of the simulated annealing procedure. cooling Cooling schedule (0 = exchange only). segments Number of secondary sampling units. controlVariable Control variable that enforces equality constraints. If a control variable is specified it must exist for each member of the population. targetVar (Required) A vector of target variances for each administrative variable. This vector must be of equal length to the number of columns of x. varWithin This parameter is used to account for within PSU variance. It is a constant value added to the target variance and PSU variance at each iteration. tolSize Acceptable difference proportion between strata sums of the control variable.

## Value

 accept Matrix with three columns and a row for each iteration, "change": change in objective function, "U": uniform random variable, "accepted": 1 - accepted, 0 - not accepted cost Objective Function label Final Label

## Author(s)

Jonathan Lisic <[email protected]>

## Examples

 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 # data set with 100 observations and four characteristics # split between two strata x <- matrix(rnorm( 100*4 ),ncol=4) # create an initial set of strata assignments label <- c(rep(0,50),rep(1,50)) #target variance targetVar <- c(10000, 20000,30000, 40000) # run equal cluster a <- saEqualClus( x, label, cooling=20, targetVar=targetVar )

jlisic/saAlloc documentation built on May 5, 2018, 6:36 p.m.