A statistician’s approach to data management

Sunday, 12 February 2012

RESOURCE CENTRE

Tax and Revenue Management: A government’s lifeblood

IT has provided the opportunities for governments to remodel the entire process of tax collection over the last decade. It is, however, a continuously evolving process and governments the world over need to constantly upgrade their tax systems to optimise their revenue workflows.

Unlocking Public Value

A recent SAP study confirmed that those organisations which adopt best practices in the areas of scope and adoption, process standardisation, technology and customer governance, do perform better, and do so as their best practice maturity increases.

Governments and Socialising

The advent of social media has seen governments hopping onto the bandwagon in a bid to further engage citizens.

Education IT

By Robin Hicks | 20 April 2010

The Centre for Research in Pedagogy and Practice at Singapore’s Nanyang Technological University is studying the relationship between the well being of students and their academic performance at school. Tan Teck Kiang, Research Associate at the centre, revealed the challenges of making sense of large, complex data sets in an interview with FutureGov.

Photos

View photos

Related Categories

From this Section

NEWS

Singapore ranks at the world number one country in educational attainment for maths, according to the Trends in International Mathematics and Science Study (TIMMS) by Boston College. Tan’s research attempts to understand the underlying factors that influence academic performance, so that the government can make informed policy decisions to stay ahead in education.

One of the survey projects consists of two assessments which track students’ scores in mathematics and English over three years of primary and four years of secondary education. Business analytics and statistical modeling were used to unearth the salient factors that might affect a student’s academic performance, such as the psychological wellbeing of the student and the demographic profile of their parents.

Before the data can be analysed it must be moulded and managed in a way that is suitable for analysis - which is never easy with social science research. “Unlike econometricians, who are interested in analysing manifest variables such as interest rates, social scientists must deal with variables that are difficult to measure,” Tan told FutureGov at the SAS Global Forum in Seattle last week.

Tan must group the data in a way that captures “unobservable variables” such as self-efficacy, using a variety of statistical models, then carry out further complex analyses using structural equation models. This is complicated as it sounds. Tan says that he cannot rely on one type of business analytics software to perform all the analyses.

“There are general statistical packages like SAS, SPSS, Statistica, Genstat, and S-Plus, among others, that cater for the general needs of statistical models. But they do not meet every need,” he says. “For very specific and well-defined statistical models, software packages like HLM, and MLWin provide alternatives, and give direct analytical output to address these modeling needs. But they do not provide general statistical support.” So he uses a combination of different platforms and packages depending on the nature of the research questions and the data.

Statistical techniques advance very quickly, and systems have to be continually updated. Keeping on top of the latest statistical analysis software makes things trickier, with new versions with new features coming out at ever-shorter time intervals. To rise to this challenge and others, Tan thinks there is a need for an international community to be formed of statisticians who use Business Analytics to share their experiences and challenges.

“Statistics is a very applied subject. With the advances in computer processing, techniques that were previously impossible, now can be done with ease,” said Tan. “To stay on top of how the discipline is changing, not only must one constantly upgrade one’s own knowledge of advanced techniques in statistics, we need a community to promote the sharing of information and best practice.”

“This would allows common understanding among the business analysts and statisticians, and help build the belief among decision-makers that business analytics produce meaningful results that improve upon the traditional approach.”