Key Messages
The Need For Advanced Data Literacy In An Age Of Accessible Statistics
In today’s world, data literacy is an invaluable skill as we are bombarded with a myriad of news stories, social media posts and arguments that use statistics as evidence.
The Art of Statistics takes this into consideration by helping you to understand how to assess the credibility of statistical figures and graphics.
It will provide you with the knowledge required to spot agenda-driven numbers or figures that don’t reflect reality.
You’ll learn how data can be used to form conclusions on topics such as whether drinking alcohol is good for your health; what remarkable creature can respond to human emotions even after it has died; and how it can also be used in forensic science to catch serial killers.
This will equip you with crucial insights into how easily misconstrued numbers can lead people astray and undermine public trust in scientific research.
Ultimately, the end goal is for you to safeguard against subtle deception and figure out who or what serves as the source for these statistical claims.
The Art of Statistics will help cast away doubt from unreliable sources and enable you to make better informed decisions when evaluating data-driven claims.
Statisticians Identify Problems, Gather Data, Analyze Patterns, And Come To Conclusions To Solve Real-World Problems
Statistics can be an important tool in helping us answer questions about the world around us.
In the case of the serial killer Harold Shipman, statistics played a key role in finding evidence that could have helped prevent his murder spree if it had been noticed earlier.
By collecting relevant data, analyzing it and examining patterns, statisticians were able to discover two major clues that indicated Shipman’s activities: he was recording a much higher number of deaths than other general practices in the area and he tended to have victims dying during times when he made home visits.
The conclusions drawn from this analysis lead to the fact that if only someone had been looking at this data and making sense of it, his murderous pattern could have potentially been identified up to 15 years before his arrest!
We can really see from this example just how powerful statistics are in helping us make better decisions – including saving lives.
Data Biases: How Context And Human Judgment Influence Our Measurements
When collecting data, it’s essential to be aware of potential biases that can affect accuracy.
Systematic bias is a common issue in data collection.
It occurs when the design or interpretation of questions leads to inaccurate results.
For instance, people may respond differently if asked ‘should we increase the voting age’ compared to being asked ‘should we reduce the voting age’.
This is why designing appropriate questions for surveys is so important.
The language used can influence how respondents answer the question and how they interpret it.
It has been shown that survey questions which are phrased in different ways can lead to drastically different responses from participants, making it difficult to draw an accurate conclusion from their answers.
In some cases, data may also be biased by what types of responses are allowed; for example, one airline survey only permitted five possible answers: ‘excellent’, ‘very good’, ‘good’, ‘fair’ and ‘ok’ – with no scope for other ratings.
This limited the level of accuracy of the results.
Due consideration must always be given to the accuracy of data which has been collected – bias can easily creep into our results if we’re not vigilant enough!
How Data Visualization And Language Frame Affect Our Interpretation Of Statistics
Data visualizations are an important tool for statisticians and communicators to present data in a clear, effective way.
This is because how data is presented can have dramatic effects on how it’s interpreted.
Statisticians often work with psychologists to determine the most effective ways of presenting data in graphical form to ensure clarity and accuracy.
A perfect example is when hospitals compare their mortality rates – the order in which hospitals are listed can give unintentional impressions about which one performs well or not, so methodical design is essential when making such comparisons.
The phenomenon of using framing to manipulate people’s emotional reactions to statistical facts has also been studied extensively.
It follows that the language used to describe factual information can be deliberately spun one way or another to evoke certain responses from readers.
For instance, saying “99 percent of young Londoners don’t commit serious violence” might reassure people about the safety of their city, but phrasing it as “1 percent of young Londoners commit serious violence” could create fear instead.
These examples show why it’s important to pay attention not only to the collection of data but also its presentation as well – how we frame statistics and design our visuals will affect how our audience interprets them and this has huge implications for policy makers and other decision-makers who must use these figures responsibly.
The Dangers Of Positive Bias In Scientific Research: Why We Shouldn’T Jump To Conclusions On Major Findings
It’s no secret that there is a positive bias in scientific literature, caused by researchers selectively reporting results to produce more publishable data.
In other words, only positive or interesting findings get reported and get published as opposed to the studies that don’t.
A great example of this phenomenon is a study carried out in 2009 which used brain imaging to study how areas of a four-pound dead Atlantic salmon’s brain would light up when it was shown photos expressing different emotions.
Of the 8,064 sites measured in the fish’s brain, 16 showed a response.
This led to sensational headline about the remarkable powers of fishes but was likely due to false positives caused by multiple testing rather than any capability of the fish itself.
False positives aren’t necessarily an issue if taken into proper context but unfortunately selective reporting means that too often it’s these positive or interesting findings that make it into scientific journals and lead to misinterpretations such as believing eating bacon sandwiches increases your risk for cancer based on one study when in reality 20 previous studies have found no link at all.
That’s why John Ioannidis famously declared “most published research findings are false” – While this wasn’t meant literally, his statement serves as an important reminder to not blindly trust everything we read without taking steps to ensure the validity and accuracy of our sources of information.
Data Journalism Is An Essential Tool For Accurately Reporting Statistics But Creative License By The Media Can Lead To Misleading Interpretations
The media often misrepresents statistics in the context of storytelling.
This can lead to inaccurate reporting about various topics and important issues.
While creating stories with an emotional punch can draw readers, it’s essential that journalists use professional judgement when deciding how to present data.
In an example related to the World Health Organization’s study on processed meat, the media reported an 18 percent relative increase in risk.
But when looking at the absolute numbers, this percentage was closer to a 1 percent increased risk – a much less dramatic one, compared to what the initial headline seemed to imply.
Therefore, it is clear that the emphasis placed on storytelling is often occurring at the expense of accuracy – which can have serious ramifications for those trying to make informed decisions based on accurate information.
The Mean Average: Why It’s Often Inappropriate And How To Spot The Difference
In The Art of Statistics, it’s shown that reported averages can sometimes be misleading when the type of average being used isn’t specified.
For example, while the mean average of 1.9999 gives us the impression that most people have 2 legs, this is actually not true since it includes those who have lost their legs in its calculation.
Similarly, a mean average of 1 testicle may lead us to believe that men and women are equally likely to have testicles which is also false as it aims to include women in the calculation.
The book then proceeds to explain the three types of average: mean, median and mode.
The mean average requires all numbers in a dataset to add up and divide by how many numbers there are; The median is the number that lies in the middle when all numbers in a data set are lined up in ascending order; Mode is the most common number in a data set.
Each form of average may be more appropriate for certain circumstances than others – for instance, if we consider an example from UK National Survey of Sexual Attitudes and Lifestyle which asks participants about their number of sexual partners – here, the mean average would skew our results due to its inclusion of outliers with much higher values than normal range (0-20).
As such, Median or Mode averages provide us with better insights into people’s typical experience compared to Mean Average.
It’s easy to take reports of statistical conclusions at face value without having an understanding of how averages were calculated.
This makes it vital for us to understand each form’s particular use cases and limitations so that we’re not misled by false reporting!
Correlation Does Not Necessarily Imply Causation: Exploring The Three Alternative Reasons For Data Correlation
Statistics is a powerful tool for understanding the data that surrounds us, but even the smartest statistician must remember the non-obvious – correlation does not imply causation.
It may seem easy to assume that if two samples of data have similar patterns then they must be related.
But this isn’t always the case, and can lead to misconceptions and misinformed conclusions.
The media can use this misconception to createconfusing headlines with false conclusions like “Why going to university increases your risk of getting a brain tumor?”.
In fact, studies like these can commonly be explained by something called ascertainment bias, meaning people from a higher socioeconomic background are more likely to get tested for and subsequently diagnosed with brain cancer.
Sometimes correlation isat best due to coincidence, like between per capita consumption of mozzarella cheese in the US and engineering doctorates from 2000-2009.
Other times it can because of reverse causation, where ill people may avoid activities such as drinking alcohol giving rise to headlines implying that alcohol consumption is beneficial.
Finallyexact correlations couldbefrom unrelated outside sources or “lurking factors”, for example during hot weather both ice cream consumption and drowning incidents rise without any causal relationship between them.
At its core, all statisticians must remember that correlation does not prove causation — period.
That’s why it’s vital to really understand the data before coming up with any definitiveconclusions; less often than we think our initial gut reactions hold true so caution needs totaken when making judgements based on datasets
It Takes Statistical Savvy To Navigate Through Probability’s Counterintuitive Choppy Waters
Probability is one of those concepts that can be tricky to wrap your head around, and it’s frequently misunderstood by many.
This was highlighted back in 2012 when 97 UK parliament members were asked what the probability was of flipping a coin twice and getting two heads – the answer being one quarter, however 60 out of the 97 parliament members could not give the correct answer.
Not only are individuals perplexed when it comes to probability, but also consider another question: an accurate mammography screening finds 1% of women have breast cancer but if a woman is diagnosed with breast cancer, it turns out she only has an 8% chance of actually having it.
This counterintuitive result again goes to show how difficult these concepts can be to grasp.
The gambler’s fallacy is also something that trips people up as they incorrectly expect outcomes from random events to align with their predictions.
But despite all this chaos there is systemic uniformity that can be observed when looking at large-scale statistics – such as suicide statistics which remain almost unchanged from year to year.
The ability for statisticians to make reliable long-term predictions about unpredictable events is remarkable and makes statistical analysis very much like ‘social physics.’
Wrap Up
The Art of Statistics by Professor David Spiegelhalter provides a detailed, yet accessible overview of the key points and findings in the field of statistics.
It outlines how statistical data can be used to help us answer important questions about our world, but it also explains how this data is at risk of being distorted or mishandled.
As such, it provides a valuable source of information on how we can improve our understanding and critical thinking when it comes to evaluating what we hear or read in the news.
The bottom line? Don’t take anything for granted: use skepticism and always look for evidence to back up any claims you see or hear.
That’s the best way to protect your data literacy!
