Evaluating multivariate optimal sample allocation tools

Overview

To support producing representative statistics the ABS conducts surveys from the Australian population of both people and businesses. A random subset/sample of the population is drawn from these populations to collect data from.

To most effectively utilise these samples the ABS uses a multivariate optimal allocation tool. This tool allocates sample to draw from subgroups of the population (‘strata’). For example, it determines how many businesses to sample for the Retail industry compared to Hospitality. This optimisation considers constraints such as cost and quality, factoring in the underlying variance in the variables of interest in these subgroups.

The ABS tool is currently maintained in SAS code created by the ABS, but are interested in moving this into R. This project would investigate off the shelf R packages to evaluate whether they could replace the existing ABS tool.

Research engagement

Critical Literature Review, Software Evaluation

Research activities

Research to understand the underlying methods, environmental scan to understand packages available, evaluation of the performance of packages compared to the ABS SAS tool.

Research skills

Data Science Literaracy, Critical Thinkin, Written Communication,

Students would gain an understanding of survey methods and conducting projects commensurate with those of a data scientist or mathematical statistician in the public service/ABS.

Outcomes

A written report and possible presentation to the ABS

The key outcome from the ABS would be an understanding of whether existing R packages could replace our existing SAS tool, and any implications from this.

Skills and experience

Data Science adn Computer Science Skills preferred

Start date

2 November, 2026

End date

19 February, 2027

Location

GP Y Block Level 8

Keywords

• Machine Learning
• Data Science
• Software Evaluation

Contact

Gentry White

3138 1658

whiteg5@qut.edu.au