Anomalously Low Variability In West Bengal’s Covid-19 Death Stats

Like most Indians, I stare daily at Covid-19 statistics, hoping for progress.

As I did so, an oddity in daily reported Covid-19 deaths stood out, specific to a particular state.

What follows is a rudimentary analysis to highlight the same.

I took daily deaths by state over last the 100-days (Aug 11th to Nov 18th), from https://api.covid19india.org/documentation/csv/.

For simplicity and ease of presentation, I limited myself to the top-10 states (by cumulative deaths over this period).

These 10 states cumulatively account for 82 per cent of all-India deaths over this period.

I ran a simple analysis of how much variability is there in this data, using two simple ratios:

• Coefficient of variation = standard deviation/mean
• An alternate coefficient of variation that doesn’t use simple-average, just in case mean was messed up by outlier data points or prior period adjustments. For this, I use Inter-quartile range (75th minus 25th percentile) divided by median

The aim of both ratios is to gauge the extent of variation around an average, for each state.

Here’s the data and one state stands out for an anomalously low variability in daily Covid-19 reported deaths:

Here’s the same in graphical form. To compare states with diverse average deaths on same scale, each state’s daily deaths are divided by that state’s average daily deaths and multiplied by 100.

So, every state’s average daily deaths is set to 100 and squiggliness of graph reflects the variability in data around a shared mean of 100.

While I refrain from making explicit inferences from this deviant data, it’s enough of an outlier to ponder about what lies beneath.