As a data-driven advisory organization, we often present complex data analyses to clients to inform their decisions. Data visualizations can play a key role in helping clients understand the context of a problem, their project’s results, potential avenues to scale-up, and beyond. Internally, we look to strengthen our tactics for communicating with various stakeholders, and to that end, regularly send around examples of our work from which to learn. This video features voices of our technical team members commenting on some of the data visualizations our team created this year and why they liked them.
Interested in data visualizations? Share with us some of your favorites in the comments — explain why you like them and why you think they do a good job of explaining complex information.
Crystal Huang: I really like this graph from the analysis of a large-scale survey we did in India of their flagship nutrition program delivered through social and behavioural change communication messaging. This is one of many data visualizations created by Nitya, Anirudha, and me. Through this program, messages about health and nutrition for pregnant women and recent mothers were delivered through 21 different platforms, which can be grouped into three broad types: 1. Digital Media, such as Facebook, Whatsapp, and TV, 2. Mid-media such as audio-visual, events, and posters and 3. Community-based events organized by community health workers. In this graph, we show exposure to any platform within each of the three broad types of platforms by socio-economic quintiles. I think this graph tells an interesting and visual story. Whereas platform exposure for mid and digital media increases with income across the quintile since these are platforms that require technology access, literacy, etc. Community engagement and inter-personal counselling can be seen as a great equalizer. The proportion of women who have attended a community event or have interacted with a community health worker is consistently 80 per cent and up, suggesting that is the platform type capable of reaching the poorest women. This finding has implications for the targeting and the social behaviour-change campaign strategy going forward.
Dan Stein: This shows a graphical representation of four impact evaluation regression results from an Orange-Fleshed Sweet Potato (OFSP) project that took place in Uganda and Tanzania. Each pair of bars shows a different regression and the way to interpret it is that the grey bar shows the control mean, while the blue bar shows the control mean plus the treatment effect. The arrow bars on the treatment are the 95 per cent confidence interval of our estimate of the treatment effect. I like showing regression results this way because you can very clearly read off the treatment effect and whether or not these effects are significant. If the arrow bar passes the control mean, that means your regressions are not significant. If the arrow bars don’t cut to the control mean, they are significant at the 95 per cent level. Sometimes people question why there are no arrow bars on the control mean and the answer is that we are not trying to compare control means and treatment means here. We are instead just trying to very clearly show the treatment effect.
1 March 2019
7 March 2019
2 April 2019
1 May 2019