|
|
Please note that all data sources for COVID-19 data have issues. There is no point asking for perfect data, you just need to be aware of the issues. For example, JHU and national sources may not agree exactly due to timing (the "day" is different in different timezones) and exactly what is collected.
Data sources
Most data downloaded daily from the GitHub repository provided by JHU Center for Systems Science and Engineering. Latest: Datafiles are here: Global cases, deaths, and recovered; Australia, China and Canada state/provincial cases, deaths, and recovered; US state/territory cases and deaths
Scottish data downloaded daily from the Scottish Government's COVID-19 page. Latest:. Datafile is here.
Victoria data downloaded daily (about 8 pm Melbourne time) from the Victoria Department of Health media updates page and public dashboard. Script help from Dave Barry. Datafile is here.
English data, including population numbers, downloaded daily from the UK Government's coronavirus data page. Datafile is here. Very recent numbers (the last 2 days) significantly undercount so are not shown - see their web page for how they collect the data.
Scottish and English regional/local data is only for confirmed cases - deaths and recovered are not available at this granularity for these countries.
Population data from a variety of sources, the main data file is here (not English regional/local data).
What is different about this webapp
There are many websites that show COVID-19 data, and some of my favourites are linked below. So what is different about this one?
This visualiser is focussed on displaying trends in the data with an empahsis on the user being able to control the visualisation and compare between countries and regions.
The nature of this pandemic is that outbreaks occur in clusters rather than being diffused through the population (at least that is my impression as an armchair epidemiologist, ie not an epidemiologist at all - my expertise is in data science and engineering). Also, the focus is on trends in the data, not the absolute or even scaled numbers. Are deaths increasing? At what rate? Is the rate increasing? How many days to double? Trends are not changed by any kind of linear scaling, although the Shape scaling is useful for seeing different countryies' trajectories.
The aim is to be able to see how well different countries, states and regions have been and are dealing with this pandemic, overall. It is NOT intended for prediction of what will happen in the future.
If you are interested in why I think "deaths days to doubling" (DDD) is a very useful metric for this purpose, look at the information on the "Views" dropdown. Or read this excellent article by Lisa Charlotte Rost.
Other sites I think do a good job of presenting some aspect of the data:
26/8/2020
Scottish data can now be viewed by Council Area
The days to doubling measure is now with reference to cumulative cases over the last 30 days - this is to be responsive to resurgence and second waves. So it is days to double the most recent 30 days total, at todays rate of increase.
14/7/2020
Scottish data now downloaded as spreadsheets rather than scraped, so this should be more reliable. Also the inclusioon of Pillar 2 data on June 15 has been pro-rata backdated. The smoothing is now a backwards-looking half-gaussian, thanks to Prof. Michael Fourman for the discussion motivating this change. And finally, the default view is now daily confirmed cases per million, smoothed 7.
8/6/2020
Added "Halftime to infection" rate metric. See Prof. Robert Rutledge's description of this metric which he invented
7/6/2020
Added sliders underneath the date scale that allows selection of the date range to display
26/5/2020
Updated with Retional and Local Government data for Victoria (Australia) and England (UK)
Updated Interface
New features include:
How do I use this webapp?
First, you can find guidance wherever you see this icon: - like the one you clicked to read this, but much smaller
Second, I recommend you look at the guidance on Trends, which will explain what this site is for, and give you links to some others that may suit your interests better
If you have any feedback you'd like to send, please direct it in the first instance to my Twitter accout
Views
This dropdown allows you to save or use views you have defined. A "view" is the settings you have chosen - geopolitical area, countries/states/regions, measure, metric, timespand, smoothing settings, and offsets if you have slid graph-lines around. Views are dynamic - when you or someone you share with loads a view, it is updated to the current data - if you want a record of the data now, take a screenshot.
Views are stored in your browser in cookies, and you are able to share them with a uniqure URL. If you share a view, it is stored on the server but no personal information about you is stored (and can't be, since you haven't entered any!). Shared views cannot be deleted from the server, but you can remove them from your browser list. You can name a view and provide a description; please do not include any personal information in the name or description. With this menu you can:
If you do not have any views stored in your browser, a default view is shown: deaths days to doubling (DDD) for the country your browser reports as being in, and the world data to use for comparison. In my view, deaths are a more reliable indicator when comparing countries, as countries tend to have fairly fixed procedures for reporting deaths, whereas reporting cases is something that can change with policy change. Days-to-doubling (DDD) is a suitable metric for comparing the trend for different countries, as this tends to be insensitive to different death-reporting standards in different countries. For those with a technical/mathematical mindset, it is essentially a measure of the slope of the logarithmic curve. Constant DDD means exponential at a rate indicated by the DDD value. There is a good discussion of DDD here.
If you do have views stored in your browser but no default is selected, then by default you will see the first one in the list (which is the first one you saved).
This controls the geographic/political areas displayed
The dropdown allows you to select one or multiple country/state/region according to the Geopolitical areas setting --- so countries or regions/territories within one country.
If you have one country/state/region selected, you will see a tooltip display of the numbers if you run your finger or mouse over the graphs.
If you have several selected, you can slide their curves left and right with your mouse/pencil/finger to try to see if other countries predict your future! The offset table on the right shows you how many days you have moved each curve.
On non-touch devices, you select one entitity by clicking on it, and additional ones using Control-Click (Windows or Linux) or Command-Click (OSX), or Shift-Click. Play around with these to see how they work, if this is unfamiliar.
The selected items move to the top of the list for easy access, and the ten most recently selected items also stay at the top of the list for easy acces
As defined in the individual datasets - some measures are not available for some zones:
Timespan for the measure - either daily or cumulative (meaning the number since the dataset started in late January 2020).
Daily data is often too spiky to be able to see the trend without some smoothing.
Different ways of displaying the same data facilitate different intuitions by your visual system
This slider allows you to smooth the daily values used as the basis for the graphs The smoothing value shown is approximately the number of days over which the data is combined. Technically we use backwards-looking half-gaussian smoothing over the prior data, with a standard deviation of approximately the number days shown on the slider. This results in about two thirds of the smoothed value being based on the current day and preceding interval shown in the slider, and one third from data prior to that (with weighting decreasing according to the gaussian distribution).
The Sliding Offsets table shows you how many days you have moved a country/region's graph forwards (right) or backwards (left) in time. To move a country's line, make sure you have checked the Timeshift checkbox above the sliding offsets table. Then use your finger or mouse to grab the line and move it. Touch and slide if on a touch device, or click and drag if using a mouse. If you want to reset a line to its original position, just unselect the country and reselect it.
These are methods for linear scaling (multiplying or dividing all values by the same number). You will notice that the rate scales (days to double and % of total) do not change with any of these adjustments - this is one reason rate scales are useful
If you check this checkbox, then you can slide each line backwards and forwards in time with your mouse (click and drag) or finger (touch and drag). You may notice that each line gets a faint thick echo around it - this is the area you need to click or touch to be able to drag. The sliding offsets table below this control will then show how many days forwards or backwards you drag each line, as you drag it. This allows you to compare lines for countries or areas where the pandemic took hold at different points in time. If you want to reset a line to its original position, just unselect the country and reselect it.
| Name: |
| Description: |
|
|
Only one view can be your default view. If you select this one, the it will be the view loaded when you visit the site again from the same browser (assuming you have cookies enabled).
If you make a view shared, then you will be shown a static URL which you can give to others for them to load this view. This URL is also available in the Views table. You can copy/paste it into an email or message. Once you make a view shared, you can no longer change the name or description, or any of its settings - it is fixed. You can set a view to be shared at any time, but you cannot unshare or delete a shared view, it will always be available at the URL provided. You can however delete a shared view from your set of browser views.