We have to take care with Covid-19 stats
16 October, 2020
Covid-19: ‘False claims of causation are frequently made by people who have an axe to grind’
• IN his letter (Take the medicine now to protect lives and economy, October 9), Professor Marc Serfaty quotes Public Health England as claiming that 3 per cent of recent Covid cases were “caused” by the hospitality industry.
If the quotation is accurate, it suggests that neither the professor nor Public Health England has much grasp of practical statistical logic.
Presumably it means that 3 per cent of people catching the virus had recently visited a pub or hotel and that was where they caught it. But it is unlikely this came from the pub staff or fittings.
It probably derived from the people they met at the pub. The cause was the desire of people (particularly the young) to be sociable, and I doubt that will be prevented by the closure of pubs or hotels.
In other words, the pub is the intermediary, not the cause; and closing pubs will not greatly affect the basic situation.
The technical name for this is autocorrelation: a situation where two series show strong correlation because both have been caused by a third factor.
For instance, most people who die have been visited by a doctor or nurse shortly before their death. The relationship is highly statistically significant. But it does not follow that (in most cases at least) the visit caused the death.
That would be due to some previous event, for instance a stroke or a road accident, which resulted in both the medical visit and the death.
False claims of causation are frequently made by people who have an axe to grind or wish to offer an apparently easy solution for a problem they cannot handle.
This has been a favourite technique of Public Health England in areas such as obesity, so it is not surprising if it is using it again with Covid. It is just sad that if its advice is taken, millions of people will lose their jobs for very little genuine health gain.
Canonbury Park South, N1