Pulling Together

 

Lies, Damned Lies, and Statistics

It was around a month ago I realised something was wrong. I was paying close attention to the daily COVID case numbers and watching for trends in the figures as I tried to understand why the average fluctuated up and down in what looked like an odd pattern.

I decided to look behind the average at the daily numbers, and that’s when I saw something worrying.

The numbers had fallen unexpectedly at the end of July and then climbed steadily all through August and the first week of September. Then they dropped back and began climbing steadily again. Why these rises and falls? Then I saw it. Although the average fell rapidly during the second week of September, that was nothing compared with the change in the daily figures. The daily count suddenly dropped overnight and then resumed its upward climb from that lower point at the same rate as before. I realised the fall was unlikely to be real. It did not seem plausible cases would really behave that way, dropping on one day and then slowly climbing as before from a lower point. No, what was much more likely was that a tranche of data had somehow dropped out of the count. I thought it was most probably a computer glitch or a laboratory’s figures no longer being included owing to a technical problem. I wanted to ask the question but was unsure to whom, and my life was rather full of other things at the time (which also accounts for these articles being much less frequent than before).

We now know what it was. A laboratory in Wolverhampton had, for reasons not yet clear, started failing to find many cases, resulting in the areas whose cases it tested getting much lower figures than the rest of the country. The Vale of Severn had particularly low numbers, and remained yellow and sometimes even blue on the BBC’s map while the rest of the contry turned redder and redder. People living there took few precautions because they believed their area largely unnaffected. Rising cases elsewhere were a mystery to us.

Once the problem was discovered, the Government decided to retest the previous week’s samples at other laboratories, and this was done over the following week. Of course, many of those samples then proved positive, but were reported a week later than they should have been, adding the same number of cases missing from previous weeks onto that week’s numbers. This inflated the figures for the week before last, causing huge media panic over the sudden rapid rise and politicians and health professionals began demanding additional precautions. No one seemed to notice the obvious fact; none of this was real!

The reason many commentators expressed bewilderment at the fall of cases at the beginning of September and the sudden acceleration in Mid-October was that neither event was true. There was no September drop and there was no rapid rise in October. Both were artefacts of the effect of the failed lab. Yet few seemed to recognise that, continuing to believe the figures as presented. Now they are puzzled that this last week they've fallen again. No, they haven’t! They never rose. That was just two week’s figures compressed into one.

This got me thinking about the average person’s ability to understand statistics. I am certainly no expert in the field. Many of the terms used in statistical analysis are a mystery to me. Yet I do know if figures go missing the result will appear lower than it should, and if extra figures are added they will inflate the results. I thought that was just obvious. Yet, it seems, less so to journalists and politicians, or maybe they’re just making mischief. Surely not!

Moving on, I then began to question the data more fully. A COVID death, for instance, being defined as any death which occurs within 28 days of a positive test result. It sounds sensible if the person concerned actually died of COVID, but when considering this can include people who die for some quite unrelated reason – other natural causes such as heart attack or cancer, accidents at work, in the home, or on the road, even if someone who had recently had COVID were murdered within 28 days they would be in the statistics – it looks less than sensible. On the other hand someone with COVID, clearly ailing from its symptoms who survives 29 days or more would not count. Suddenly, it is apparent that deaths within 28 days are not necessarily linked to the disease at all, even if one assumes the majority might be. A terminally ill patient diagnosed with COVID fewer than 28 days before they succumb will count as a COVID death.

This led me to think about the claim people are more likely to die within 28 days if they are elderly or unwell. That would actually be true even without COVID. A person who is elderly or unwell is more likely to die than a person who is well. In fact, out of 100,000 80 year olds, 50,000 of each sex, the expected deaths within any 28 day period would average in normal times 391 (1,374.5 men and 1,166.7 women pa), according to the government’s own data at https://www.ons.gov.uk/file?uri=%2fpeoplepopulationandcommunity%2fbirthsdeathsandmarriages%2flifeexpectancies%2fdatasets%2fsingleyearlifetablesuk1980to2018%2fsingleyearlifetablesuk/singleyearlifetablesuk.xlsx. How then did the risk per 100,000 drop as low as 117 for unvacinated and about 45 for vaccinated over-80-year-olds after a positive test? The answer, I surmise, is that this is not per 100,000 80+s, but per 100,000 of an all-age population. Without knowing the number of 80+ people in the 100,000, we can draw no information about the impact of COVID in terms of excess deaths within the period, because we cannot deduce the number of people who would have been expected to die anyway. The information therefore becomes meaningless because it is impossible to interpret. It gives only an impression which might, or might not, be true.

Such are the statistics on which we are expected to base our decisions, and by which the Government submits its policy to scrutiny. They really are not good enough for that purpose. One piece of information clearly lacking to create a context for it would be the numbers of people who die within 28 days after a negative test. That would at least give us a basis for comparison.

About the Author

K J Petrie has a Full Technological Certificate in Radio, TV and Electronics, an HNC in Digital Electronics and a BA(Hons) in Theological Studies.

His interests include Christian and societal unity, Diverse Diversity, and freedoms from want, from fear, of speech, and of association. He is a member of the Social Democratic Party.

The views expressed here are entirely personal and unconnected with any body to which he belongs.

Engage with the Author

If you’d like to discuss anything please send me your e-mail address and I will send you mine.

Your address will only be used for replying and will not be passed to anyone else.

Subscribe

If you would like to be informed by e-mail of new Pulling Together articles as they are published, please enter your address here.

Your address will only be used to let you know about new articles and will not be passed to anyone else.

Full List