Premium
Understanding relationships between global health indicators via visualisation and statistical analysis
Author(s) -
Lodha Suresh,
Gunawardane Prabath,
Middleton Erin,
Crow Ben
Publication year - 2009
Publication title -
journal of international development
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.533
H-Index - 66
eISSN - 1099-1328
pISSN - 0954-1748
DOI - 10.1002/jid.1652
Subject(s) - bivariate analysis , visualization , bar chart , scatter plot , bivariate data , data science , computer science , statistical graphics , data visualization , causality (physics) , health indicator , public health , data mining , statistics , machine learning , graphics , mathematics , medicine , computer graphics (images) , nursing , physics , quantum mechanics
Several agencies such as World Bank, United Nations and UNESCO are disseminating a large amount of socio‐economic data at national level. Various websites such as UC Atlas, Gapminder, CIESIN and NationMaster are attempting to provide general users visualisation tools to display this data. Typical visualisation methods include line graphs, bar graphs, scatter plots, colour‐coded glyphs (such as circles) and world maps. In addition to the general public, there is great interest in educational, research and public policy institutes to try to understand the relationships between these socio‐economic indicators. In this paper, we juxtapose two techniques to investigate the relationships between global health indicators. The first approach employs sophisticated statistical techniques to develop a causality model between various global health indicators. The second approach, typically employed by the visualisation users of the various websites mentioned above, is to utilise a bivariate display between the health indicators in order to discover relationships between these variables. This visualisation approach is perhaps closest to a bivariate regression or correlation. Therefore, we employ these simple statistical techniques and associated visualisations as well. In this work, we analyse the two approaches using two specific examples related to health indicators. We find that the two approaches sometimes agree strengthening the conclusions or may provide different perspectives that require more careful analysis of the conclusions and need for further research. Copyright © 2009 John Wiley & Sons, Ltd.