June 5, 2014

NZ interactive graphic examples


  • From The Wireless, a story with maps of voter turnout and registration rates for younger people (RadioNZ might not be where you expect interactive graphics, but there it is). If I were being picky, I would say the popup labels are too big relative to the size of the map window.

Thomas Lumley (@tslumley) is Professor of Biostatistics at the University of Auckland. His research interests include semiparametric models, survey sampling, statistical computing, foundations of statistics, and whatever methodological problems his medical collaborators come up with. He also blogs at Biased and Inefficient See all posts by Thomas Lumley »


  • avatar
    Richard Penny

    Would be useful to have some indication of a credible interval for the the numbers. Particularly since people are going to compare numbers. Even reading the report appendices doesn’t tell you much.

    3 years ago

    • avatar
      Thomas Lumley

      Very true.

      3 years ago

    • avatar

      Thanks for the comment and I hope in the future we can be better on this issue. I’m Peter Ellis, Manager of the Sector Performance team that worked with Dragonfly to produce the interactive web tool for the Regional Economic Activity Report.

      There’s three interesting challenges that stopped us showing uncertainty. One is the constraints of the data visualisation medium itself. It’s non-trivial to show uncertainty with choropleth maps and treemaps. One cool idea is to have a probability distribution for each statistic and then animate the graphics to show the range of those statistics – either systematically pulsing through the range, or just doing simulations. There’s no way we could have done this in the time and resources available, and I doubt we could in 2015 either, but I throw the idea out there for someone who wants to build on it for your own applications! I haven’t seen anyone do this to show uncertainty but it might work quite nicely.

      But the second problem is the difficulty with producing those credibility or confidence intervals in the first place. This is often difficult for official statistics because of the complexity of the estimation process. You virtually never see confidence intervals for national accounts data for example, which is the source of the Regional GDP estimates (http://www.stats.govt.nz/browse_for_stats/economic_indicators/NationalAccounts/RegionalGDP_HOTPYeMar13/Data%20Quality.aspx). The tourism spend data we used was based in part on the International Visitor Survey (see http://www.med.govt.nz/sectors-industries/tourism/tourism-research-data/international-visitor-survey/about-ivs/data-reliability) and we could in principle have propagated the sampling error from that survey into the country estimates, but in the production of the regional tourism estimates a range of non-sampling errors must also be introduced (for example, by raking to national accounts product estimates; and by relying on electronic transactions for regional allocations) and they are impossible to quantify. Similarly the census data behind the household income and demographic data is subject primarily to non-sampling error and hence not easily quantifiable.

      The third challenge is general statistical literacy levels in the public, and unrealistic expectations of precision that come about in part from the narrow confidence intervals possible when only taking into account sampling error and easy to measure variables, often binary or a few simple choices, like voting intention. The media tend to be highly critical of the width of confidence intervals on other variables when we draw attention to them, particularly on hard to measure things such as spend and income. I’d welcome ideas on how best to engage with the public on statistical uncertainty.

      None of this is to argue that we shouldn’t be doing more on drawing attention to uncertainty, so I hope we can move more in this direction in coming years.

      3 years ago