Books like Thinking About Statistics

In the ongoing debates over artificial general intelligence (AGI), Judea Pearl is taking a firm stand: He argues that an intelligent robot should be able to reason about causality and that the currently fashionable approaches to AI miss this aspect.

A celebrated AI researcher and a Turing Prize laureate, Pearl has developed an amazingly original approach to this problem. This book is a high-end popular exposition of his approach.

But it’s so much more than that. It’s a history of statistics and its conflicted attitude to causality. It’s a story of heroes (or villains?) in this history. And it’s a scientific autobiography that describes Pearl’s journey. Pearl likes picking fights with the AI community, statisticians, or economists. He’s boastful, provocative, extremely intelligent, and knows how to tell a story.

As I write these lines, artificial intelligence (AI) is getting a lot of attention.

This is largely due to ChatGpt recently bursting onto the scene. But even before ChatGpt began making its mark, AI was often in the news. Some have expressed worry that it will take our jobs, others that it will reinforce systemic oppression by making racially or otherwise discriminatory decisions, and some have even voiced concerns that one day a superintelligent AI might pose an existential threat to humanity.

In the midst of all this, what might get lost is what AI is, what it's capable of doing, and what its limitations are. Wenger's book is intended to address all of these questions. It manages to do so in a way which goes into some of the mathematics of AI systems and yet remain accessible to a lay audience.

After laying out the technical aspects of AI,… show more.

Many quant geeks are familiar with statistics. The dominant school of statistical thought is called "Frequentist" or "Classical."

It focuses on either 1) testing a given hypothesis by determining how likely observed data are on the assumption that the hypothesis is true or 2) constructing intervals for which a certain percentage of them contain the actual value of whatever is being estimated.

A lesser known, although this seems to be changing, school of thought is Bayesian statistics. It focuses on using prior information about some phenomenon in order to revise or update one's beliefs about it.

If you're into stats but don't know much about Bayesian statistics, Donovan and Mickey's book is a great place to start. It's somewhat mathematical but covers the technical aspects much more accessibly that any other book I've seen on the topic. 

Many people associate mathematics with calculating things or plugging numbers into formulas to get answers to a multitude of problems.

But this isn't how mathematicians view their discipline. They see mathematics as more about starting with definitions of key mathematical concepts, stating axioms about these concepts, and proving things about them. For those interested in going from calculating and plug and chug mathematics to "real" mathematics, Richard Hammack's book is a terrific place to start.

The book covers a number of topics that cut across all of pure and applied mathematics, topics such as sets, relations, and functions. But the heart of the book is focused on how mathematicians go about proving things. If one wants a glimpse of how mathematicians really work, go out and get this book immediately.  

All politicians should be forced to read this book. Anyone who reads a newspaper should be forced to read this book. My favourite radio programme in the world is Tim Harford’s More or Less. And this book is every bit as good. Harford is clear, incisive, and always interesting. In a world crowded with disinformation and fake news, he shows you how to evaluate the numbers that are thrown at you. To read him is to become a little cleverer. Make this man prime minister someone.

I love this book because it reminds us of the many ways that technology can affect justice.

It is tempting to think sophisticated tactics such as “predictive policing” can solve all problems relating to human bias. However, Brayne shows that data and algorithms do not eliminate bias and discretion. Instead, high-tech police tools simply make bias less overt and visible, which erodes the public’s ability to hold the police accountable.

I especially enjoyed how the book flips the script, considering diverse ways to use these tools to help the public. For example, how can municipalities use technology to analyze the underlying factors that contribute to policing problems in the first place?

I learned Tukey’s name about as soon as I learned that data visualization existed as more than a menu in Excel and a personal obsession. Tukey coined the term “exploratory data analysis,” and so tapped into a passion for swimming around in all the interesting rows and columns. Tukey was working before computers were widespread, and so I got a view of how he saw data: working against the constraints of pencil and paper; keeping your hand moving as fast as possible. While the explorations we can do with gigabytes of memory and powerful rendering are very different, the goal of getting information into your head as fast as possible is unchanged.

This book applies economic theory and econometric methods to problems in criminology.

It is divided into three parts. Part I discusses models of criminal recidivism. Part II describes the economic model of crime. Part III estimates cost functions for prisons.

Specific chapters cover statistical analysis of qualitative outcomes; analysis of two measures of criminal activity – the arrest rate and the conviction rate; and long-run estimates of cost functions for a group of Federal Correctional Institutions. 

I am laughing out loud, even now that I am rereading this book for the umpteenth time. Fraudsters are so clever, and so is advertising. And then there is sloppy journalism with its “wow” statistics.

I like his book enormously, not least because of its witty illustrations. It is subversive, comic, and provocative, and it makes me wise to seductive, misleading practices–and it does so with a light touch.

From well-known authorities in the R-sphere (including a former R Core Team member), this is a long-standing text whose first edition was one of the early books intended to teach R to beginners. It provides concise instructions and examples on how R is used as a programming language before focusing on 'number-crunching' statistical methods that are typically seen as computationally intensive. One of the notable features of this book is the statistical methods at hand are not just illustrated using 'black-box' code--the reader is provided with the necessary mathematical detail to understand what's going on behind the scenes for those that are so inclined.