Bayesian Emulation and History Matching of JUNE

Abstract

We analyse JUNE: a detailed model of Covid-19 transmission with high spatial and demographic resolution, developed as part of the RAMP initiative. JUNE requires substantial computational resources to evaluate, making model calibration and general uncertainty analysis extremely challenging. We describe and employ the Uncertainty Quantification approaches of Bayes linear emulation and history matching, to mimic the JUNE model and to perform a global parameter search, hence identifying regions of parameter space that produce acceptable matches to observed data.

Visualization for epidemiological modelling: challenges, solutions, reflections and recommendations

“…They should be addressed through a combination of applied visualization work in epidemiology and elsewhere, and controlled experimental approaches that complement this activity. In short, we need to establish, develop, assess and refine… effective and efficient ways of visualizing differences in quantities that allow us to make spatial and temporal comparisons [ 155 ]: 788 (see also [ 17 ], fig. 2 for a temporal view of a spatio-temporal dataset); methods for comparing quantities and ratios that vary by orders of magnitude [ 159 ]: 2380 (see also [ 82 ]) and that control for population size [ 155 ]: 788 ; effective ways of using layout and colour in combination in dense data graphics [ 139 ] (see also [ 17 ], fig.…”

“…In short, we need to establish, develop, assess and refine… effective and efficient ways of visualizing differences in quantities that allow us to make spatial and temporal comparisons [ 155 ]: 788 (see also [ 17 ], fig. 2 for a temporal view of a spatio-temporal dataset); methods for comparing quantities and ratios that vary by orders of magnitude [ 159 ]: 2380 (see also [ 82 ]) and that control for population size [ 155 ]: 788 ; effective ways of using layout and colour in combination in dense data graphics [ 139 ] (see also [ 17 ], fig. 4); visualization idioms to deal with large numbers of data items and new structures in data that are unexpected or important [ 160 ]: 705 , [ 150 ]: 687 ; consistent visual languages —something that is hard to achieve in a pandemic (in parallel)—that allow us to use colour, icons, other encodings and even interactions in ways that are common, predictable, consistent, effective and understood [ 153 ]; narrative patterns for communicating in cases where subjects are sensitive or controversial [ 152 ], [ 153 ]: 706 ; approaches that minimize misinterpretation and account for it where it occurs [ 155 ]: 788 by addressing some of the open issues listed above, and through effective documentation, signposting, training, learning and co-design processes; effective ways of further embracing the emerging digital workspace for long-term immersive visualization support; and reliable and effective processes for conducting and supporting research through applied visualization that draw upon the themes identified through this engagement between epidemiological modellers and visualization researchers (§3e).…”

“…effective and efficient ways of visualizing differences in quantities that allow us to make spatial and temporal comparisons [ 155 ]: 788 (see also [ 17 ], fig. 2 for a temporal view of a spatio-temporal dataset);…”

“…effective ways of using layout and colour in combination in dense data graphics [ 139 ] (see also [ 17 ], fig. 4);…”

“…7), [ 7 ] (fig. 4), [ 83 ] (figs 4 and 5), [ 17 ] (figs 5 and 6). Knowledge about how to use visualization effectively in any applied context draws upon these established visual solutions and their use to address particular problems (combinations of task, dataset and user) [ 80 ].…”