Visualizing the trends in confirmed Covid-19 cases in the UK
Bar plot
Here I’ve used a barplot to depict the counts of confirmed Covid-19 (UK) cases reported at: https://coronavirus.data.gov.uk/?_ga=2.167663289.643846418.1587577545-996826653.1586878543#countries
If the level of testing associated with each daily count of confirmed cases is known, the resulting graphical trends would be more informative/useful. If the number of tests, for example, on Sundays are lower, this could be reflected on the daily count of confirmed cases for that day.
Overall we can see that the daily number of new confirmed cases is plateau-ing.
Different trends would be expected for daily confirmed cases, hospital admissions and deaths.
The data used in the abovementioned plots refer to confirmed cases on a given day. I’ve used the daily reported cases to construct the cumulative data series.
Time series
Here the trend in the count of confirmed cases of Covid-19 (UK) is depicted in line graphs. Data reported at: https://coronavirus.data.gov.uk/?_ga=2.167663289.643846418.1587577545-996826653.1586878543#countries
The data used in the abovementioned plots refers to confirmed cases on a given day. Also, the interventions highlighted do not include the huge financial investments made the and new hospitals constructed to tackle the pandemic. As of April 18th 2020, for example, over £3.2 Billion has been allocated to councils; and as of April 20th 2020 the Covid-19 Job Retension Scheme (CJRS) was announced.
Visualizing the trends in cumulative Covid-19 deaths in the UK
Trend in deaths (hospital)
Below, I’ve plotted the trend of cumulative Covid-19 deaths on a log scale (instead of on a linear scale) so that it is easier to detect deviations from an Exponential progression; in particular, making it easier to detect a slowing down of the progression. There were missing values for days 41 and 43. The daily reported deaths were obtained from: https://coronavirus.data.gov.uk/?_ga=2.167663289.643846418.1587577545-996826653.1586878543#countries and I’ve used the daily reported deaths to construct the cumulative data series.
In the following plot, I’ve sliced the abovementioned plot (with accompanying loess smooth) and added a grid to illustrate the behaviour of the slope of the curve at various slices.
Finally, I’ve plotted the log cumulative deaths for England, Northern Ireland, Scotland, Wales and the United Kingdom. This is followed by the daily hospital deaths for each region. The orange line in these two plots represents a trendline obtained from locally estimated scatterplot smoothing. This is done by performing local regression models using subsets of data surrounding each point of interest. The choice of neighbourhood for constructing each data subset is informed by a prescribed nearest neighbour algorithm.
The data source is: https://coronavirus.data.gov.uk/
Social distancing, spatial interactions, and risk of infection
– musings—
The UK social distancing policy recommends that individuals maintain a spatial distance of 2 metres apart from each other. For each individual, let \(x_{i}\) be associated with a zone of influence \(Z_{i}\), such that \(Z(x_{i},r)\) is a disc of radius r = 1 metre, centered at the x-y location of each individual \(x_{i}\) and that \(Z(x_{i},r)=\left\{a\in\Re^{2}:\left\|a-x_{i}\right\|\leq r\right\}\).
The zone of influence (as in Area Interaction Point Process models) can be thought of as the region within which an individual is able to exert an impact on another in a spatial context. This concept has traditionally been used in modelling the spatial locations of plants/animals and the zone of influence in the case of plants, for example, would primarily reflect the impact an individual plant can have on the soil nutrients available to another neighbouring plant.
An interesting extension to the definition of the concept of a zone of influence is such that it represents the region(“infectious zone”) within which an individual with Covid-19 can potentially infect someone else.
Assuming that the “true” zone of interaction is 1 metre, the area of overlap could be an important measure in tracking the spread of Covid-19. It would be essential, however, to account for the fact that within the zone of influence, the risk of infection is lessened as the distance from the center of the zone increases.
For two individuals \(x_{1}\) and \(x_{2}\), 2 metres apart from each other, and each with a zone of influence of radius 1 metre, the area of overlap between their zones of influence would be zero. Of course, an immediate challenge would then be to decipher whether super-spreaders would have a larger zone of influence/infection than other spreaders and to decipher whether the “true” zone of influence has a radius of 1 metre.
The illustration in Figure 1 represents xy location of 12 individuals; the shaded light blue regions denote the area of overlap between each individual’s zone of influence with that of nearest neighbours. Only one individual observes the social/spatial distance and hence the area of overlap of this person’s zone of influence with that of neighbouring individuals is zero. The red circles represent infected air borne water droplets within the zone of influence of an individual with Covid-19 .