Much of my work has been on developing interactive statistical graphics for high-dimensional data, and the implementation has been in these software packages: xgobi, ggobi, cranvas. The primary methods include tours, projection pursuit, manual controls for tours, pipelines for interactive graphics, a grammar of graphics for biological data, and visualizing boundaries in high-d classifiers. I have also experimented with visualizing data in virtual environments, and found that people do see clusters better in that environment than on a single computer screen.

My current work focuses on bridging the gap between statistical inference and exploratory graphics. We are doing experiments using Amazon's Mechanical Turk, and eye-tracking equipment. We have found that we can crowd-source people to read plots that can provide statistical significance on visual discoveries. Its very exciting work. We can also use the crowd-sourcing methods to rigorously test whether one data visualisation design is better than another for communicating information.

Some of the applications that I have worked on include backhoes, drug studies, mud crab growth, climate change, educational testing, gene expression analysis, butterfly populations in Yellowstone, stimulus funds spending, NRC rankings of graduate programs, technology boom and bust, election polls, soybean breeding, common crop population structures, insect gall to plant host interactions, bushfires, soccer and tennis statistics. I am currently looking at Melbourne's pedestrian sensor data. One of the things that I recently found is that boys do NOT universally do better than girls on average on the PISA math scores, and that a handful of countries including UAE, Jordan, Qatar, Thailand and Malaysia have a reverse gender gap - girls score better on average than boys. This finding appears to have been picked up and entered into wikpedia. However, girls universally score better than boys on average in reading.