Data visualization for development
6 likes1,507 views
The document is an introductory guide to data visualization for beginners, emphasizing the importance of storytelling and asking good questions in data science. It covers the data collection process, tools for data visualization, and considerations for effective communication of results. The guide also provides resources for further learning and highlights the need for data cleaning and understanding audience context.
Data visualization for development
- 1. Data Visualisation for Beginners Sara-Jayne Terp 2nd October 2014
- 2. Introduction (XKCD.com: data science humor)
- 4. Not Talking About âą Algorithm details âą Machine learning âą Big data techniques âą Koalas Cogs image from Worradmu on freedigitalphotos.net; Koala from koalastothemax.com
- 5. Data Visualisation Images: Hans Rosling Ted Talk; erm.. something about India and women
- 6. Data Science
- 7. The Process
- 8. Storytelling
- 9. Ask Good Questions âData science is all about asking questions. You engage in it whenever you interactively and iteratively search for deep, hidden patterns.â â James Kobielus âą Do people have more phones than toilets? âą How is Ebola spreading? âą Is using wood ïŹres sustainable here? âą Can we feed 9 billion people? (Simple, Actionable,Incremental)
- 10. Get the data
- 11. An Aside: Big Data
- 12. Volume
- 13. Velocity
- 14. Variety CSV, json, xml, excel, pdf, text, webpages, rss, scanned pages, images, videos, audiofiles, maps, proprietary formats etc.
- 15. Veracity The â3 other Vsâ are Viability, Validation, Verification. ! Validation, checking that inputs are real, is a big deal for development data.
- 17. Explore your Data âą Spend time with your dataset: âą Understand where it came from - can you live with the assumptions the data collectors made? âą Look at it âą Plot it âą Where are there holes? Inconsistencies? Anomalies? âą Clean your data, ïŹnd better datasets, get more data
- 18. Data is Often Inconvenient (plus often sitting on someoneâs laptop)
- 19. Data is (almost) Always Dirty DR Congo in data.UN.org: âCongo, Democratic Republic of theâ, âCongo Democraticâ, âDemocratic Republic of the Congoâ, âCongo (Democratic Republic of the)â, âCongo, Dem. Rep.â, âCongo Dem. Rep.â, âCongo, Democratic Republic ofâ, âDem. Rep. of Congoâ, âDem. Rep. of the Congoâ ! DR Congo in common standards: âDemocratic Republic of the Congoâ (UN Stats), âCongo, The Democratic Republic of theâ (ISO3166), âCongo, Democratic Republic of theâ (FIPS10, Stanag), â180â (UN Stats), âCODâ (ISO3166, Stanag), âCGâ (FIPS10)
- 21. Process your data Everything is a dataset (if you look hard enough)
- 22. Process your data (The relationships between things are interesting - these are my Facebook âfriendsâ, on Gephi)
- 23. Process your data (machine learning can be useful too, e.g. if youâre working in a language with no stopword lists)
- 24. Model your data âą Youâre persuading people with âtruthsâ: do your best to make sure theyâre truthful âą Always cross-check âą Statistics is your friend
- 25. Explain your results âą Youâre trying to persuade people to change: âą Their opinions âą Their actions âą Visuals are (often) more persuasive: âą âI already knew that increased incarceration didnât lower crime, but I wasnât sure of the statistics. To see it on the graphs is really eye opening.â * *: Pandey et al, The Persuasive Power of Data Visualisation
- 28. Tools: QGIS
- 29. Tools: Google Fusion Tables (link to google spreadsheets)
- 30. Tools: Tableau http://www.tableausoftware.com/public/gallery/major-league-baseball
- 31. Tools: Tableau (this is a choropleth)
- 32. Tools: Python, R (AFINN sentiment analysis: sometimes you have to code)
- 33. Tools: D3, Javascript D3 gallery: https://github.com/mbostock/d3/wiki/Gallery
- 36. Whatâs a dashboard? (Indiaâs government employee attendance system)
- 37. Psychology checklist âą Whatâs your important message (and what are you trying to hide?) âą Medium matters (laptop, phone, sms?) âą Colours matter âą As do angles and relative lengths. âą And think about your audience, e.g. what local effects do you need to be aware of, how do you compensate for colorblindness etc etc. âą New visualization type? Check the Gestalt principles http://www.creativebloq.com/how-design-better-data-visualisations-8134175.
- 38. Please donât do thisâŠ
- 39. Tools checklist âą Whoâs your audience? âą Whatâs the medium: paper, static webpage, tablet, phone? âą Which languages do you need to display? âą And are they right-to-left? âą Is this a one-off visualisation or will you need to update it as new data comes in? âą Are your audience viewing this online or ofïŹine? âą What resources do you have for updating the visualisation?
- 40. Where to go from here âą Websites, e.g. Information is Beautiful, DataScience Central, ïŹowing data, ILoveCharts, Chart Porn, junk charts, visual.ly blog, ïŹvethirtyeight.com âą Meetups and events, e.g. DataKind, NYC Data Skeptics âą Books e.g. Nathan Yao âVisualise this!â, anything by Tufte âą Spring course on data science
- 41. Ask good questions; Tell good stories