Authors: Boyana Buyuklieva*, Centre for Advanced Spatial Analysis, University College London, Adam Dennett, Centre for Advanced Spatial Analysis, University College London
Topics: Spatial Analysis & Modeling, Migration, Communication
Keywords: Geo-computation, Migration Metrics, Reproducible Research
Session Type: Paper
Start / End Time: 3:20 PM / 5:00 PM
Room: Bayside A, Sheraton, 4th Floor
Presentation File: No File Uploaded
Advancing any field requires a common foundation of reproducible research. Arguably the increasing growth of recent research utilising geo-compuational techniques in urban analytics has been facilitated though the easy sharing of data and methods though tools like github. However, with camps emerging characterised by researchers favouring one or another language, challenges remain if all research is to be easily reproducible by everyone, regardless of their tools or versions of choice. Being explicit about methods is especially important in interdisciplinary research, where stakeholders, by definition, come from different backgrounds and with different assumptions and conventions. To explore these issues, this paper uses the substantive topic of methods for understanding migration and residential mobility processes in England and Wales focusing on four metrics types after Bell et al. 2001: migration intensities, connectivity, system impact and distance. We present a framework for communicating methods and metrics consistently and open the discussion on reproducible research in computational geography. Migration measures will be used to discuss and reflect on pseudo-code and literate programming as a means of setting a common baseline for communicating geo-computational metrics that is neither too programming language specific, nor too vaguely described. The former is problematic because it is not robust enough to guard against changes or version compatibility in the tools. The latter issue is important, as different implementations of formulas, such as where parenthesis are set or rounding at different steps, can produce different results, making studies inconsistent and difficult or even impossible to compare.