Visual Data Science: Incorporating statistical modelling into visual exploratory data analysis

Elicit requirements, design, implement, test and evaluate visual encodings and interactions for supporting explorative geographical data analysis with appropriate statistical and data mining techniques. These techniques should support analysts using domain knowledge to build simple models on-the-fly, gradually refining them using various model fit metrics and summary statistics, with the goal to understand the processes and drivers that led to the original data. Part of this process is maintaining the provenance of which models were built and their characteristics. Restricted to geographic data with features that are generally understood by analysts (i.e. not high dimensional data)


  • Review: Exploratory visual interfaces, analytical "goals" in these interfaces, suitable statistical and data mining techniques, their application

  • Design: Based on one or more case studies, design visual encodings and interactions to support exploratory data analysis with appropriate statistical and data mining techniques

  • Implement: Implement versions of this design in order for them to be test and evaluated

  • Test and evaluate: assess the impact this has on the exploratory visualisation process

  • Provenance: Maintaining the provenance of which models were built and their characteristics and make available to support the data exploration process    


  • Development: ability to build interactive visualisation interfaces that connect to statistics and data mining libraries/software

  • Statistics: a good understanding of applied statistics

  • Statistical graphics: knowledge of the use of statistical graphics

  • User-centred work: open to designing, testing and evaluating these interfaces with analysts from different domains  

  • Other: Handling datasets, working with geographical data


This topic was suggested by Dr Aidan Slingsby. Please direct further enquiries to the giCentre.