Statistical errors and a wild night on the tiles

·3-min read
<span>Photograph: Adrian Sherratt/Alamy</span>
Photograph: Adrian Sherratt/Alamy

Carlo Rovelli makes a profoundly important proposal (Statistical illiteracy isn’t a niche problem. During a pandemic, it can be fatal, 26 October) and focuses particularly on the need for schoolchildren to understand how statistics work, and their relevance and applicability. It would be possible to include in such teaching the equally important concepts of scale and magnitude; these are just as powerful in testing the accuracy of the picture of the world which we are given.

So often there is a failure to grasp that assertions (by politicians, for example) can be easily tested. The size of an entity (for example, the numbers of those defrauding the benefits system) or the relative sizes of two entities (say, the degree by which the wealth and health of the rich exceed the wealth and health of everyone else) can be easily estimated by logical steps, easily taught.
Greg Conway
Amersham, Buckinghamshire

• Having profoundly disagreed with the likely results of the throwing of 100 grains of rice at 400 tiles suggested in Carlo Rovelli’s article – “We tried numerous times … and there was always a tile with two, three, four, even five or more grains on it” – and not having 400 tiles, I numbered some imaginary tiles from one to 400 and instructed Excel to do the rice-throwing (100 random numbers between one and 400) and then count how many tiles were occupied not at all, once, twice, thrice, etc. On average, it was just under 312, 78, 10, one respectively, and thereafter none. I calculated that were I to repeat the experiment 20 times or so, I might be 50/50 for a four-occupancy, but for five, that would be 400 times.

Clearly with the table in the middle, not to mention unspecified numbers of people occupying otherwise perfectly good tiles, the basic assumptions of this experiment were violated. But I’m wondering how long this party went on for. And who did the clearing up?
Caroline Matheson
Teaching associate, Selwyn College, Cambridge

• We should all be taught statistics, but reading and interpretation are also important. The example given in Carlo Rovelli’s article is wrong. We are told that the disease in question is “rare” and “non-infectious”. The “rice-throwing” example then given equates to a common disease: one that will affect 100 people spread across 400 workplaces. Other important factors are not given. How big is the workplace? If five people in my department (of around 50 people) came down with the same rare, non-infectious disease, I would indeed be looking for common factors in my workplace, which would be correct.
Prof Scarlett Thomas
University of Kent

• Hooray for an article on statistics – perhaps all your columnists could read it. With all the figures quoted about Covid, there has been not one serious statistician involved who has explained their reliability. If all politicians were forced to quote their margin of error, we would perhaps have fewer people believing that the whole thing is a hoax or conspiracy.
Mo Whiteford
Malmesbury, Wiltshire

• Carlo Rovelli is spot-on about the need to understand statistics better. Could journalists stop talking about relative risk and start talking instead about absolute risk? If the risk of something is one in a million and the risk doubles (“Eating quinoa doubles risk of X disease, shock horror!”), this means it shifts to one in 500,000, which is still extremely small.
Roger Fisken
Reading, Berkshire