.mau. wiki

This page is read only. You can view the source, but not change it. Ask your administrator if you think this is wrong.
===== Sanne Blauw -The Number Bias =====

=== Foreword: Captivated by numbers (p. 9) ===
=== 1 Numbers can save lives (p. 15) ===
=== 2 The dumb discussion about IQ and skin colour (p. 29) ===

▪ p. 46 - These kinds of interpretations are dangerous. If you want to take figures seriously, you should acknowledge that there is a great deal that they do not say. In other words, that GDP is merely a measure for ‘production’ and IQ no more than a score on a test. Instead, due to convictions and biases, the figures are inflated into something they are not.

=== 3 What a shady sex study says about sampling (p. 49) ===

▪ p. 53 - Two years after Kinsey’s report had been published it was down to the three statisticians to assess whether he had done a good job. Their quest exposes six crucial mistakes that can be made when using samples. 1. The circumstances or questions are flawed; 2. The survey excludes particular groups; 3. The interview group is too small (The random sample, a solution to the problem?); 4. Too few people want to take part; 5. The margin of error is overlooked (you can look up an online calculator on a website such as goodcalculators.com, which works out margins for random samples); 6. A particular outcome matters to the researcher

=== 4 Smoking causes lung cancer (but storks do not deliver babies) (p. 67) ===

▪ p. 67 - It was 1953 and the tobacco industry was in trouble.1 Shares in Philip Morris & Co., United States Tobacco Company and other manufacturers suddenly plunged in value. The immediate cause was a publication by cancer researcher Ernest Wynder and colleagues, who had painted tar from cigarettes onto the shaved backs of white mice using a camel-hair brush.

▪ p. 69 - ‘Big houses attract big, and potentially big, families,’ Huff writes, ‘and big houses have more chimney pots on which storks may nest.’

▪ p. 71 - 1. It’s a coincidence: [...] In 2010, Paul the Octopus predicted the results of eight World Cup matches. Time and again he used his tentacles to open the correct food box, the box with the flag of the football team that would win the subsequent match.

▪ p. 71 - analist Tyler Vigen. He became famous as a result of the strange correlations he publicised on his website Spurious Correlations.16

▪ p. 72 - this cartoon [https://xkcd.com/882/] shows two additional, prevalent problems in science. The first is publication
bias. [...] Only 5 per cent of a coincidence!’ By this, cartoonist Munroe is referring to the so-called p-value, which measures to what extent the outcome came about as a result of coincidence. The renowned statistician Ronald Fisher was responsible for ensuring that, during the twentieth century, the p-value became the method to measure the significance of a correlation.

▪ p. 73 - The p-value is the likelihood that, in case the jelly beans do not actually cause acne at all, you will still end up finding a certain higher percentage of acne patients in the jelly bean group. If this likelihood is below the agreed threshold – often 5 per cent – then the likelihood that you will detect this percentage of patients is so small that you can call the correlation ‘statistically significant’.

▪ p. 75 - 2. A factor is missing. [...] ‘The point is that when there are many reasonable explanations, you are hardly entitled to pick one that suits your taste and insist on it.’

▪ p. 76 - 3. There is a reverse causal relationship 

▪ p. 77 - NOS had overegged the message a bit, too: that ‘almost twenty times’ should have been ‘almost 20 per cent’ [..] But an important detail was missing in much of the reporting: 20 per cent of what? If you look at the data, six in a hundred Dutch people will get bowel cancer at some point in their life, According to the World Health Organisation, this percentage drops by 18 per cent – this is where that ‘almost 20 per cent’ came from – if you stop eating processed meat.

▪ p. 80 - In 1979, the Tobacco Institute, an institute financed by the tobacco industry, published a graph showing the development of different types of cancer. [...]  The graph was supposed to show that this was not necessarily the case. It gave a picture of the proportion of patients with mouth and throat cancer, bladder cancer and cancer of the oesophagus. The result looked so messy that it was hard to argue there was a consistent increase. But what was missing in the graph? Sure enough, the most important effect of smoking: lung cancer.

▪ p. 82 - On 14 December 2015 the National Review, a conservative American magazine, tweeted: ‘The only #climatechange chart you need to see.’ The image showed the temperature since 1880. The outcome? The average temperature had barely changed over the past 135 years. The line showing temperature change was as flat as the heart monitor output of a patient who’s just died. 

// Il trucco qui è la scala scelta //

▪ p. 86 - In it, Huff describes three types of, what I like to call, ‘cocky correlations’ – correlations that pretend to be something more than they are: causal relationships.

=== 5 We should not be too fixated on numbers in the future (p. 90) ===

▪ p. 92 - The assumption that we can be oblivious and let numbers make decisions about our life is dangerous. Behind this notion lurks a serious misunderstanding: namely that the data always corresponds with the
truth 

ricordarselo sempre. Il problema non sono (solo) i dati errati


=== 6 Our psychology decides the value of numbers (p. 107) ===

▪ p. 107 - although a high mortality risk had been found among beer drinkers, in wine drinkers it turned out to be minimal. It wasn’t so much the alcohol, Prasad suggested, but the lower income of beer drinkers that was unhealthy. 

//questo vale per l'Olanda, però.//

▪ p. 108 - But the longer I wrote about number misuse, the more I began to doubt whether knowledge was the only solution.

▪ p. 108 - It’s often easy to recognise these errors by asking a few questions. How was the data standardised? How have the figures been collected? Is there a causal relationship?

▪ p. 108 - I wished the ground would open up and swallow me when, after a lecture, I saw that 50 per cent of those attending had not rated my performance as good. But I forgot to take into account that only two people had taken part in the survey.

▪ p. 109 - In numerous instances in this book, researchers were influenced by their – conscious or unconscious – biases and convictions. And we number consumers are equally prone to this.

▪ p. 110 - Our brain works like a lawyer; it will find arguments
to defend our convictions, whatever the cost.

▪ p. 111 - So when you see a number, take a step back and ask yourself: what do I feel? When I saw the alcohol study mentioned above, for example, I became irritated. Especially when I later read the headline ‘An extra glass of alcohol can shorten your life by 30 minutes’.14 This was simply total nonsense. My irritation was a feeling that matched my professional ‘tribe’ – number sceptics – but also my personal one. When I meet friends, we drink a few glasses of wine or beer. That’s what we do. Should I stop doing this? I’d rather not. I felt pleased when I read the tweets from the renowned Vinay Prasad. Relieved; I could carry on drinking.

▪ p. 112 - In a follow-up experiment Kahan presented respondents with two articles about climate change; one that
confirmed the concerns about it, another that was sceptical. The headline of one of the articles had been worded in such a way as to appear surprising: ‘Scientists Report Surprising Evidence: Arctic Ice Melting Even Faster Than
Expected’. The other article appeared to be reporting nothing that was new: ‘Scientists Find Still More Evidence that Global Warming Slowed in the Last Decade’. Which article do you want to read, he asked? And this is where he discovered the power of curiosity. Curious types did not choose the article with the headline that accorded with their convictions, but the challenging one. For these respondents, curiosity was a stronger force than ideology. 

//devo dire che anch'io avrei scelto così.//

▪ p. 112 - This experiment is educational. If you encounter a number, don’t stop and just accept it, but go and explore. Search – on- or offline – for people who look at the number from a different angle. Don’t only read articles that confirm what you already think, but look for information that may make you feel uncomfortable, angry or desperate.

▪ p. 114 - But people who are certain, by definition lack curiosity. If you hang on to your convictions at all cost, you are never receptive to new information.

▪ p. 114 - Numbers cannot answer that question for you. They
can seem like the ideal excuse to stop thinking, but they cannot provide quick
and easy answers. At best, they will help you navigate the terrain.

=== Afterword: Putting numbers back where they belong (p. 116) ===

▪ p. 116 - Take GDP. Over the past few years, unease has begun to surface about the limitations of GDP and the dominant role it plays in relation to government policy. Various measures that could replace or complement GDP have been suggested. Some countries, for example, now measure their citizens’ happiness;1 the OECD created the Better Life Index, a broader indicator that takes into account factors like the environment or the job market in a particular country;2 and Statistic Netherlands (CBS) has recently started measuring ‘general concept of well-being’, which, among other things, studies the effects of our prosperity on future generations.

▪ p. 117 - Since 2012, economists and other social science researchers have been registering their experiments with the American Economic Association before they actually conduct their research.4 This means it’s immediately clear what they are planning to do, so that they do not endlessly look for significant results later on.

▪ p. 118 - Take the OpenSCHUFA initiative.11 SCHUFA is the largest credit bureau in Germany. Their credit scores have major consequences for the financial situation of an individual, but the company refuses to make their algorithm public. However, according to German law you, as a citizen, can request your own report. In 2018, Open Knowledge Foundation and Algorithm Watch, in fact, called on German citizens to apply for their credit reports and forward them on. With sufficient data, they would be able to reverse engineer the algorithm. Within a few months, more than 25,000 people had requested their own credit report. 


=== Checklist: What to do when you encounter a number (p. 119) ===

1. Who is the messenger? 2. What do I feel? 3. How has it been standardised? 4. How has the data been collected? 5. How has the data been analysed? 6. How have the numbers been presented? An average / A precise figure / A ranking / A risk / A graph