Introduction to dot charts and forest plots

Dot charts can be used to display any grouped data values including frequencies, probabilities, proportions, and any outcome measures by group. Since it is a very simple plot type, dot charts can be easily understood by many but are found aesthetically pleasing by few. Therefore such visualisations are more likely to be used with a specialist audience such as regulators and other experts. A variation of the dot chart is the forest plot which contains more statistical underpinnings and can be used to communicate summary measures such as mean risk difference and risk ratios as well as their associated uncertainty (via confidence intervals). Forest plots can be used as means of benefit-risk communication to specialist audience such as physicians, the regulators and other experts. A dot chart is very similar to a bar chart, and it has been argued that it is a better alternative to a bar chart since a dot chart has very high data-ink ratio, that is, it embraces simplicity by presenting only the crucial data points (see Tufte 2001). On the other hand, some experimental evidence suggests dot charts are not perceived as well as bar charts when it comes to information extraction. Dot charts have only one value axis, whilst the other only represents discrete entities such as groups (similar to histograms). Therefore users must recognise that the dots should not be joined and that they are not used to represent the relationship or variability in the data. Dot charts can be produced in software packages such as Stata, R, SAS, SPSS, Tableau, Spotfire, QlikView, IBM Many Eyes, Microsoft Excel, and many others. Forest plots can be produced easily in statistical software packages such as Stata, R, and SAS, and may be produced with a bit more work in software such Microsoft Excel and Tableau. Interactivity may be limited for dot charts, but may be more suitable for forest plots. An interactive forest plot may allow users to explore different levels of benefit-risk criteria and to input their own preferences into the underlying model. An example of an interactive forest plot created in the Natalizumab Wave 2 case study.