World Statistics Day only comes every five years - like the Olympics - so it's time to express a little mindful gratitude for all the statisticians and other wonderful nerds out there helping us to understand the world.
Since I literally wrote the book on the topic, I’d like to mark the day by sharing my TEN RULES FOR THINKING DIFFERENTLY ABOUT NUMBERS. Each of us could be thinking more clearly about the world if we got ourselves right with the numbers. timharford.com/books/worldadd…
So, Rule One: SEARCH YOUR FEELINGS.
What we believe, or refuse to believe, is strongly influenced by our emotional reaction. A lot of the statistical claims we see aren’t just data: they are weapons in an argument. Social media thrives on emotion. So do media headlines.
I don’t think we can or should ignore our emotions or values. But we should get into the habit of noticing them. If the reaction to a claim is a knee-jerk, “this proves I was right!” or “Fake news.”, we need to count to three and start thinking more clearly.
Rule Two: PONDER YOUR PERSONAL EXPERIENCE. We get information from graphs and spreadsheets. We also get information from the rich and vivid experiences all around us. To adapt Kahneman, there are “statistics, fast and slow”.
Often our personal experience conflicts with the data. Sometimes the data are misleading. Sometimes it’s our own experience that is skewed in some way. One thing I learned from the great @HansRosling was that the best insights come when we are able to combine the two.
Rule Three: AVOID PREMATURE ENUMERATION. One of the perils of being numerate is the temptation to start chopping up numbers – computing ratios, rates of increase, means and variances – before we understand what they refer to.
Arguably the financial crisis of 2007-08 was caused, in part, by very sophisticated mathematical analysis of measures of risk which were poorly understood. Start by understanding what the numbers refer to, eh?
(Sorry for the break. Just had to pop onto @BBCr4today to beat the drum for WORLD STATISTICS DAY.)
Rule Four: STEP BACK AND ENJOY THE VIEW.
Rather than focusing on the latest data point, get some context. What has happened to this data series over the past year? Past decade? What is happening elsewhere? Are there any comparisions which help make sense of the number?
For example - here, @standupmaths and I ask how many dinosaurs to the bus? (And other vital questions.)
Rule Five: GET THE BACK STORY.
Every number is in front of your eyes for a reason – often because it is particularly surprising. It may be surprising because it is not representative of a wider trend. It may be surprising because it is flat-out wrong.
The “replication crisis” in psychology came about as scholars realised that peer-reviewed work was subject to a powerful “interestingness” filter – good-but-dull work was buried, while flukes were published. But that same powerful filter, publication bias, is everywhere.
Rule Six: ASK WHO IS MISSING. You’ve all read @CCriadoPerez 's “Invisible Women” (I hope) – about how the data we gather often miss the needs of women, or fail to disaggregate so that we can ask questions about the different experiences of women and men. amzn.to/31nTf1C
And once you start thinking like that, you see data gaps everywhere – from the shy Trump voters who pollsters missed at the last election, to grand big-data based claims that only measure those of us with smart-phones.
Rule Seven: DEMAND TRANSPARENCY WHEN THE COMPUTER SAYS ‘NO’
Algorithms are providing some amazing insights – but very often they are commercially confidential.
Senior managers & politicians swallow tall tales about the power of the algorithm (eg “this algorithm will give everyone the perfect A-Level grade”).
But those grand claims are often unproven. We should demand better evidence and independent scrutiny.
In “How To Make The World Add Up” I write about the history of alchemy vs science. Same methods, same people, different results: largely because science involved collaboration, open experimentation and scrutiny. Alchemy was all done in secret.
Rule Eight: DON’T TAKE STATISTICAL BEDROCK FOR GRANTED
And this is the real message of World Statistics Day. All around us, statisticians and other hero-geeks are painstakingly gathering and analysing the numbers we need to understand what is happening all around us.
The pandemic provides a vivid example: how does the virus spread? Where is it now? How dangerous is it? Who is most at risk? What sort of settings or behaviours increase the risk? What treatments are safe and effective?
Such questions can’t be answered with statistics.
But we’re trying to answer other questions – about crime, the economy, demographics, health, the environment and more – all the time. The statistics are quietly and carefully assembled and we take them for granted.
We just assume they will always be there when we need them. We notice only when something goes wrong. That is a shame. So let us celebrate the wonderful folks who help us see the invisible all around us.
Rule Nine: REMEMBER THAT MISINFORMATION CAN BE BEAUTIFUL TOO
The hero of my chapter nine is Florence Nightingale – someone who really understood the power of data visualisations to win an argument.
But there is a risk in data-visualisations. Alberto Cairo points out that our visual sense is so potent we use phrases such as “I see” as a synonym for “I understand”. Very often, however, the understanding is illusory – we’ve been fooled by beauty. amzn.to/2IMmXqC
All the other rules apply, with double force, when looking at a picture of data.
Rule Ten: KEEP AN OPEN MIND
Easy to say, isn’t it? But what I mean here is a willingness to keep examining the data and being able to admit to yourself and others that you have changed your mind.
Irving Fisher, one of the tragic heroes of Chapter Ten, was a genius – but failed to change his mind when it mattered. He lost everything.
A rival, not without sympathy, said that Fisher’s undoing had come because “he thinks the world is ruled by figures instead of feelings”.
I hope this thread has persuaded you that in fact, the world is ruled by both.
I should add: all too often, statisticians are prosecuted, persecuted, threatened, or in extreme cases executed - all for speaking truth to power.
Some of these tales are in "How To Make The World Add Up".
A sobering piece in @signmagazine here:
BONUS! FREE EXTRA GOLDEN RULE!
The Golden Rule is...
Be Curious.
The world is a fascinating place, but many of the things that make it fascinating are people too numerous ever to meet, patterns to subtle to detect unaided, or other truths that only statistics can reveal...
...so have a little confidence in your ability to make sense of the numbers, and show some curiosity. Ask what's going on in the world - and how numbers can help to illuminate it.
THREAD 1/ "When the facts change, I change my opinions. What do you do? - Attributed (without evidence) to John Maynard Keynes, hero of "How To Make The World Add Up" ch 10
Why is it so hard for people to change their minds?
2/ Partly, we make public statements and then we get stuck. We feel don't want to admit making a mistake. Opponents call us out for our inconsistency. A shame.
3/ But it should be really easy to update beliefs based on new information. For example, I wrote in August that the chance of being infected was 44 in a million per person per day. I still believe that is true.... of August.
Source: ons.gov.uk/peoplepopulati…
My column this morning ventures into science fiction: what if everyone who was infectious glowed orange like the children in the Ready Brek ads?
The answer: the virus would be extinct in humans within a month.
This, basically, is the promise of super-fast, super-cheap testing: test everyone, all the time, and the problem goes away (as @paulmromer said many months ago).
A few problems, though:
a) We don't have billions of rapid tests, and as @deeksj reminds us the testing industry is long on promises and short on solid evidence.
b) Boris Johnson has said it will happen - so obviously it won't.
c) Cheap tests will be ropey and unreliable.
Statistics, lies, and the virus: five lessons the pandemic has taught us about data and how we use it.
My #LongRead for the @FTMagft.com/content/92f64e…
Lesson One: the numbers matter.
We've become used to numbers being spun, distorted, used for slippery targets, lied about - and we easily become cynical. But statistics aren't just a vector for bullshit: they're the only hope we have of understanding the pandemic.
Lesson Two: don't take the numbers for granted.
Even nerds like me can easily lapse into thinking that statistics just come from some big database somewhere. But first they have to be gathered, measured, collated etc. This 'statistical bedrock' is essential, and under-rated.
1/ Time for an apology and a correction. Seems that every newspaper in the UK is (correctly) reporting that I said the risk of catching a fatal case of Covid-19 is about the same as the risk of having a bath. I did say that, but I was wrong. Details below.
2/ What’s true is that for a typical 60 year old, running the sort of infection risks the current UK citizen is currently running, the chance of catching a fatal case of Covid-19 is currently about 1 in 2 million per day, perhaps a bit lower.
3/ Now according to this piece – the author of which should be held blameless – the risk of taking a bath is about 1 in 3 million (0.3 micromorts). But that can’t be right. theconversation.com/whats-most-lik…
1/ So, a few words about my friends at @MathsGear - the most wonderful online store for nerds everywhere.
MathsGear is the brainchild of brilliant maths & science performers @standupmaths@MouldS@jamesgrime@stecks - who have adapted brilliantly to evaporation of audiences...
2/ Where else but @MathsGear can one buy non-transitive dice? (Yellow beats red, red beats green, green beats blue, blue beats purple... purple beats yellow...) mathsgear.co.uk/collections/di…
3/ And where else but @MathsGear can one buy skew dice, which look weird but are actually perfectly fair? (Hurry: these sell out fast.) mathsgear.co.uk/collections/di…
A friend of mine thinks it's too risky to go out, and asked me to do the maths for him. I did: ft.com/content/176b9b…
You need to make a lot of assumptions and guesstimates - too many. I wish the data were better (@amymaxmen has a great recent piece on this in @NatureNews). But there is enough information to make educated estimates.
In the UK, according to last week's ONS population survey, there were 22-76 infections per million people per day in England, with a best guess of 44. If you behave like a typical English resident for one day (no, I don't know what that means either) then...