cipriancraciun
commented
4 years ago I have myself generated a few heatmaps based on the JHU dataset:
- https://github.com/cipriancraciun/covid19-datasets/tree/master/plots/jhu/countries
- https://github.com/cipriancraciun/covid19-datasets/tree/master/plots/jhu/subcontinents
- for others see https://github.com/cipriancraciun/covid19-datasets/tree/master/plots/jhu
And indeed there are "strange" datapoints that stand out. (BTW, similar behaviour can be seen with the ECDC and NY Times datasets, also available at the links above.)
However I would say that the major culprit here is poor reporting internally at country level, and then poor reporting to the public.
In fact I thought of applying some "smoothing" that would keep the total number intact, but "spread" the values to neighboring days.
No one has recovered in the US according to this data. https://github.com/CSSEGISandData/COVID-19/tree/master/csse_covid_19_data/csse_covid_19_time_series
-- nothing here.
inconsistent. death toll, infected. inconsistent and incomplete. https://github.com/CSSEGISandData/COVID-19/tree/master/csse_covid_19_data/csse_covid_19_time_series confirmed only globally and china.
use of a heatmap to represent numbers is misleading.