Towards explanatory model building in social data science

Talk presented at RGS-IBG Urban Analytics Workshop, 31st August 2016

Transcript

1. Improving analysis of passively collected data through probabilistic analysis Roger Beecham, Cagatay Turkay & Jo Wood giCentre, City University London
2. Towards explanatory model building in social data science Roger Beecham, Cagatay Turkay & Jo Wood giCentre, City University London Improving analysis of passively collected data through probabilistic analysis
3. Can we use bikeshare data to learn (new things) about urban cycling behaviour?
4. Geographic Data Science Ÿ Computational Social Science Ÿ Social Data Science Ÿ Empirical Sociology Ÿ Data-driven Social Science Ÿ Urban Informatics Ÿ Urban Analytics Ÿ …
5. Geographic Data Science Ÿ Computational Social Science Ÿ Social Data Science Ÿ Empirical Sociology Ÿ Data-driven Social Science Ÿ Urban Informatics Ÿ Urban Analytics Ÿ … Repurposing new, population-level data for social research
6. o  Measurement validity •  Are we observing what (the behaviours) we think we are? o  Incomplete data •  We don’t know enough about users Challenges
7. o  Measurement validity •  Are we observing what (the behaviours) we think we are? o  Incomplete data •  We don’t know enough about users Challenges ----------------------------------------------------------- Confirmatory / explanatory analysis?
8. Research ambition Take a research finding (observed cycling behaviour). … And provide explanations for that finding. … With quantified measures of each explanation’s effect.
9. Exploring gendered cycling behaviours within a large-scale behavioural data-set Transportation Planning and Technology, 2014 Vol. 37, No. 1, 83–97, http://dx.doi.org/10.1080/03081060.2013.844903 Exploring gendered cycling behaviours within a large-scale behavioural data-set Roger Beecham* and Jo Wood Department of Computing, giCentre, City University, London, UK Transportation Planning and Technology, 2014 Vol. 37, No. 1, 83–97, http://dx.doi.org/10.1080/03081060.2013.844903 2015
10. Studying commuting behaviours using collaborative visual analytics Roger Beecham a,⇑, Jo Wood a, Audrey Bowerman b a giCentre, Information Sciences, City University London, United Kingdom b Delivery Planning - Cycling, Transport for London, United Kingdom a r t i c l e i n f o a b s t r a c t Computers, Environment and Urban Systems xxx (2013) xxx–xxx Contents lists available at ScienceDirect Computers, Environment and Urban Systems journal homepage: www.elsevier.com/locate/compenvurbsys
11. Geography of scheme provision and usability
12. mid point least quiet routes Provision of cycle- friendly infrastructure most quiet routes
13. Research ambition Explain the observed geography of bikeshare (estimated) workplaces | given |
14. Research ambition Explain the observed geography of bikeshare (estimated) workplaces | given | •  Geography of jobs available London and accessed by the same home locations (resident population) of bikeshare users [2011 Census]
15. Research ambition Explain the observed geography of bikeshare (estimated) workplaces | given | •  Geography of jobs available London and accessed by the same home locations (resident population) of bikeshare users [2011 Census] •  Provision of bikeshare cycling infrastructure [LCHS usage data]
16. Research ambition Explain the observed geography of bikeshare (estimated) workplaces | given | •  Geography of jobs available London and accessed by the same home locations (resident population) of bikeshare users [2011 Census] •  Provision of bikeshare cycling infrastructure [LCHS usage data] •  Provision of “cycle-friendly” routes [cyclestreets.net]
17. observed bikeshare jobs at lsoai =
18. observed bikeshare jobs at lsoai = available jobs at lsoai | lsoai...n homeplaces of bikeshare commuters [Census 2011]
19. observed bikeshare jobs at lsoai = available jobs at lsoai | lsoai...n homeplaces of bikeshare commuters [Census 2011] + bikeshare infrastructure at lsoai [Density of docking spaces, LCHS | competition, Census 2011]
20. observed bikeshare jobs at lsoai = available jobs at lsoai | lsoai...n homeplaces of bikeshare commuters [Census 2011] + bikeshare infrastructure at lsoai [Density of docking spaces, LCHS | competition, Census 2011] + provision of “cycle-friendly” routes to lsoai [Quietness scores from estimated routing data, cyclestreets.net ]
21. Comparing geography of bikeshare workplaces men’s vs women’s Standardised residuals for studying lsoa-level variation Cramer’s V for studying effect size across lsoas = obslsoai explsoai p explsoai c = s P 2 N(k 1)
22. to 0.
23. Standardised residuals for studying lsoa-level variation Cramer’s V for studying effect size across lsoas Comparing geography of workplaces bikeshare vs modelled = obslsoai explsoai p explsoai c = s P 2 N(k 1)
24. Assumptions dataset available jobs at lsoa Cramer’s V: 0.24 Cramer’s V: 0.25 men women
25. Cramer’s V: 0.18 Cramer’s V: 0.20 Assumptions dataset available jobs at lsoa + capacity men women
26. all available jobs all available jobs | bikeshare lsoas all available jobs | bikeshare lsoas + quietness all available jobs | bikeshare lsoas + capacity Developing a model (quantified explanations)
27. Model development Re-evaluate groundtruth (estimated workplaces) Investigate other explanatory variables. Better modelling infrastructure. Locally-varying (gw) regression coefficients for certain explanatory variables. Policy related outcome -- workplace potential? Parts of London with high labour market demand, poor supply of ‘infrastructure’ – bikeshare and general cycling. Further considerations
28. roger.beecham@city.ac.uk gicentre.net/rogerbeecham/ Thanks
29. all available jobs all available jobs | bikeshare lsoas all available jobs | bikeshare lsoas + quietness all available jobs | bikeshare lsoas + capacity