Stratified Path Sampling

Abstract

This paper proposes the employment of path sampling when the population is stratified into smaller regions, called stratified path sampling, resulting in more precision estimation. An estimator of the population mean and its variance including the variance estimator are provided. Simulation study was performed to examine the efficiency of stratified path sampling compared to stratified random sampling by using real-world population data. The benefit and disadvantage of stratified path sampling are described. From the simulation result, stratified path sampling is less efficient in non-rare population with low variation of study variable values, but it is more efficient in rare population and non-rare population with high variation of study variable values, particularly when the starting and ending units are on the columns containing large values of study variable.