Friday, April 17, 2020

17/4/20: COVID19 Updated Charts and Outliers


Updating two charts for #COVID19 pandemic today:

First: US vs EU chart:

Second: Russia chart:

Since I included no commentary on Russian data in the chart itself, it is worth noting that data so far indicates no data suppression or mis-reporting. This is confirmed by analysis of 'outliers' in the data. I have looked at all countries with > 1,000 cases reported and considered observations on cases reported that fall out of trend line from the time when the country cumulated cases counts reached > 50 cases. For example, if a country reported 127 cases in day T, followed by 139 cases in day T+1, and suddenly showed 0 cases in T+2, followed by 99 cases in T+3, the date of T+2 was marked as an 'outlier'. I ignored all cases where 'outlier' suspect dates were above 20 cases, even if the number was still outside the range of the trend-defined 'norm'.

Note: these outliers can be a function of tests arrivals dates, availability of tests, hospitals reporting dates and other differences that have nothing to do with 'Government manipulation'. All in, 43 countries out of 77 with more than 1,000 cases have reported at least one outlier.

Russia had 3.45% of days reporting appearing as extreme outliers. 30 out of the total 77 countries on the list had higher percentage of outliers days than Russia. Median for 77 countries was 2.9%, mean was 5.9% and STDEV was 8.6%.

Only two of these countries, namely Russia (3.45% of observations countable as outliers) and China (13.3% of observations being outliers), has been accused in the Western media of releasing politically manipulated data. China, of course, has a very high percentage of observations that can be identified as outliers, while Russia is, basically, middle-of-the-road.

No comments:

Post a Comment