### Introduction to statistical maps

Statistical maps are used to show the difference in values (frequency of an event, probability of an event etc.) between different geographical regions in geo-spatial analysis. Modern statistical graphics include use satellite images and street-level views. Statistical maps may not be very relevant for use in the benefit-risk assessment. However, it could be used to illustrate geographical variations of certain events such as the differences in prevalence of events, and benefit-risk decisions in different regions. The use of statistical maps in areas such as weather warnings and predictions ( UK Met Office) is common, and so the general public, patients, physicians, regulators and other experts may be familiar with format.

Some knowledge of the geographical regions, and any additional coding, displayed on the map would be required to understand statistical maps. In geo-spatial visual representation, the information on physical proximity (for benefit-risk of medicines) in general may not always be useful (see Few 2009). Statistical maps however do not always pinpoint the exact location of an event, which may lead to misinterpretation. The use of colours on maps to represent values can also reduce the accuracy of the quantitative data (see Cleveland 1984).

A different type of statistical map is the "sector map" which is also known as a "treemap" and a "mosaic plot". Sector maps are based on statistical summary of various measures, and plotted as rectangles of sizes proportionate to the magnitude of the measures, and often colour-coded. However, this type of representation may be affected by the limitations of area judgment and colour. Sector maps requires more statistical knowledge in the construction to be accurately interpreted, but could be used by most audience as investigative tools.

Statistical software packages like Stata, R and SAS support geo-spatial mapping. Tableau, Spotfire, QlikView, IBM Many Eyes and Google Drive also support the production of geo-spatial mapping. SAS and R software packages could integrate with Google Maps as discussed in http://cran.r-project.org/web/packages/RgoogleMaps/vignettes/RgoogleMaps-intro.pdf for R. Interactive use of maps could have features such as highlighting areas, tooltips for annotations, filters by events, and ¡®zoom in¡¯ to show more details on the geographical region such as street names. A function to fade the map into the background may also be convenient to bring forward the quantitative data shown on the map (see Few 2009).