Thursday 28 May 2020

Day 73 of self-isolation - Moving averages

Moving averages

Each weekend, there is a dip in the numbers (also on holiday Mondays). The reason for this dip, I don't know, but I'm quite sure that it isn't the virus taking a couple of days off. It surely has to do with the reporting system, and the fact that although rapid response to medical issues are the top priority of every doctor, the filling in of statistical forms are not.

The issue of the weekend dip in the numbers, can be handled in one of two ways. The first way is to use a seven day moving average; the second way is a technique called "seasonal adjustment", that allows for a regular variation in statistics around the seasons of the year - in this case, it would be applied to daily data with a seven day cycle.

I chose to use a seven day moving average. Seasonal adjustment has two problems. First, it's really best applicable to a situation where the underlying data is stable or slowly rising or falling. The pandemic numbers are obviously not like that. The second reason is that it is a lot more difficult to explain how the calculation is done, whereas a seven day moving average is self-explanatory. If anyone needs it to be explained, please ask.

The seven day moving average also deals with the problem of "holiday Mondays", which also appear to have a strong influence on the daily numbers. One disadvantage of the moving average, is that it will always lag three days behind the daily report. That can't be helped, but the moving average line (in red on my charts) and be extended "by eye" if you want to.

The data come from the usual source - the worldometer.com numbers. I've extracted the numbers and created a Libreoffice spreadsheet to make the charts. As usual, I'm offering the spreadsheet to anyone who wants to play with it.

So here I present the numbers for the UK; cases and deaths. The day by day numbers, and the seven day moving average.

As you can see, the case numbers are falling,but are still 1/2 of the level that they were at the peak. Death numbers are also falling, and are a third what they were at the peak.

In my view, the numbers haven't gone down to a level that makes reopening safe. Maybe by June 15 they will be low enough.