Improved policymaking one data set at a time

From poor to excellent, how would you rate your health last year?

Roy Costilla points at his research poster

Doctor of Philosophy in Statistics candidate Roy Costilla is developing methods to find patterns and hidden groups from responses over a number of years to questions like these that will one day allow for better modelling and improved policymaking.

This method of estimating patterns and hidden groups from data is known as a cluster analysis of respeated ordinal data, a method that Roy is also applying to other areas such as genomics and bioinformatics.

“It’s been a journey learning to work with finite mixtures and Bayesian models—the mathematically complex models we are applying to estimate patterns in repeated ordinal data. The possibility that our research can help uncover latent groups that we don’t currently see in raw data is really exciting.”

Part of a Marsden-funded group at Victoria University of Wellington for cluster analysis, along with supervisors Ivy Liu and Richard Arnold, Roy started a collaboration with Religious Studies academics at at Victoria University of Wellington and Psychology academics at Auckland University who have supplied him with new ordinal data from the New Zealand Attitudes and Values Survey.

“This comprehensive data set is a longitudinal data set made and designed for New Zealanders and run by New Zealanders, so I’m excited about the future benefits it might have for our communities in terms of more robust policymaking.”

Roy has been supported with funding for research costs and opportunities to go to conferences and training workshops representing Victoria University of Wellington throughout New Zealand and in Australia, Mexico, the US and Peru.

Roy says that students have access to top computer and library resources. “All the knowledge is here. There is almost no limit to the resources you can get in the Library and there are several grid computing clusters that have dedicated support from Unix experts at the School.”