Dr Simon Rudkin, University of Manchester
Data visualisation is a core component of the empirical process, from data exploration to the evaluation of models. However, visualising datasets with many continuous variables is challenging. To see the shape of data, scatter plots (and their derivatives) are the go-to graphics. Considering each variable as a dimension on the page, we are limited to two directions, with other variables needing to be captured in size or colour. Many solutions to viewing multi-dimensional data involve loss of information in dimensionality reduction, or combine many plots of subsets of variables.
Topological Data Analysis Ball Mapper is a means to create an abstract two-dimensional visualisation of multivariate data without loss of information. A motivation for the method, outline of the algorithm, and look at the research agenda developing the method are provided. This seminar explains why we all need to consider the structure of our data and place data visualisation at the core of our empirical analyses.
This is a precursor to a methods@manchester workshop on Topological Data Analysis Ball Mapper, which will run later in the Autumn.
