Some of the examples, like the route cause of the 2008 financial crisis and the Monty Hall problem, have been widely telegraphed I d assume Wheelan s explanations will make them easier to sink in I took statistics classes during two phases of my education, but am currently using little of it at work Given big data is on the rise, and large free data sets are becoming obtainable, I m toying the idea of taking data crunching as a pastime.

To challenge myself, I opted for the audiobook and played it at double speed Not until double digit chapters when I had to pause and rewind, because keeping four digit numbers in my head became impossible on the noisy subway That said, Wheelan did not lose a beat, or his cool, in breaking down difficult hypothesis testing and multivariate regression for a lay person I applaud making audiobook available for math related subjects it would benefit visually impaired students who might otherwise find math s too daunting On that note, Jonathan Davis did a wonderful job with his smooth narration and sense of humor pretty much syncs with the author.

I couldn t get through this book, mainly because I know too much about statistics and I know too much about the specific examples he gives to illustrate his points Unfortunately, while at times Wheelan does convey the underlying concepts of probability and statistics in a way that would help you understand them at a basic level, he does so in what I would regard as a patronizingly oversimplified way If you compare this book to Nate Silver s book on prediction or, indeed, to the book he says motivated him How to Lie with Statistics , this book simply doesn t deliver the goods It clothes the concepts of statistics in yet another layer of misunderstanding and half truth If, for example, he had spent a chapter on unemployment and really showed how, as a descriptive statistic, the number is meaningless for all kinds of measurement and theoretical reasons, I would have been impressed Instead, he used it as an example of a good statistic If he had cited Savage s The Flaw of Averages while making points about averages, dispersion, and distributions the wrong points, I might add , I would have been impressed If he had at least mentioned Bayes Rule and Bayesian statistics, I would have been impressed I wasn t impressed.

I have already talked about statistics here, and not in good terms It was mostly related to Nicholas Nassim Taleb s works, The Black Swan and Antifragile But this does not mean statistics are bad They may just be dangerous when used stupidly It is what Charles Wheelan explains among other things in

**Naked Statistics**

**Naked Statistics**belongs to the group of Popular Science Americans often have a talent to explain science for a general audience Wheelan has it too So if you do not know about or hate the concepts of mean average, standard deviation, probability, regression analysis, and even central limit theorem, you may change your mind after reading his book Also you will be explained the Monty Hall problem or equivalent Three Prisoners problem or why it is sometimes better even if counterintuitive to change your mind.

Finally Wheelan illustrates why statistics are useless and even dangerous when the data used are badly built or irrelevant even if the mathematical tools are correctly used Just one example in scientific research which is another topic of concern to me This phenomenon can plague even legitimate research The accepted convention is to reject a hypothesis when we observe something that would happen by chance only 1 in 20 times or less if the hypothesis were true Of course, if we conduct 20 studies, or if we include 20 junk variables in a single regression equation, then on average, we will get 1 bogus statistically significant finding The New York Times magazine captured this tension wonderfully in a quotation from Richard Peto, a medical statistician and epidemiologist Epidemiology is so beautiful and provides such an important perspective on human life and death, but an incredible amount of rubbish is published.

Even the results of clinical trials, which are usually randomized experiments and therefore the gold standard of medical research, should be viewed with some skepticism In 2011, the Wall Street Journal ran a front page story on what it described as one of the dirty little secrets of medical research Most results, including those that appear in top flight peer reviewed journals, can t be reproduced If researchers and medical journals pay attention to positive findings and ignore negative findings, then they may well publish the one study that finds a drug effective and ignore the nineteen in which it has no effect On top of that, researchers may have some conscious or unconscious bias, either because of a strongly held prior belief or because a positive finding would be better for their career No one ever gets rich or famous by proving what doesn t cure cancer Dr Ionnadis a Greek doctor and epidemiologist estimates that roughly half of the scientific papers published will eventually turn out to be wrong Pages 222 223 This is not the most exciting book ever, but it s way exciting than you would think for a book about statistics More importantly, people YOU NEED TO KNOW THIS STUFF This is how you separate the lies from the damn lies from the nonsense that TV news shows spew at you I don t care if you

**read**THIS one, but please just fucking

**read**a book about statistics THANK you.

Simplify ,

Reading the book helps you become critical so you won t naively believe a person or organization s argument when they cite statistics to support their case or when you

**read**about scientific breakthroughs in the newspaper or other claims based on statistics.

There are many popular science

**books**that try to teach basic statistical concepts, but often than not they fall into the awful popular science trope of narrative over concepts that Malcolm Gladwell introduced into science writing and then Jonah Lehrer perfected into an awful, horrible art Take Nate Silver s lauded book The Signal and the Noise Each chapter is about some specific area of prediction, and along the way some statistical concepts are introduced but rarely elaborated I will note that Nate Silver only rarely mentions what the expert had for lunch during their interview, unlike much worse science

**books**that presume we are interested in the culinary habits of scientists In that book, Silver also tries to make a case for Bayesian statistics over traditional statistics, but because the explanation of the concepts is not very rigorous, we don t get so much an argument as an opinion.

Charles Wheelan s book is a fantastic antidote to modern popular science writing and conceptual hand waving In a nutshell, the book is a stats 101 course without the math Unlike, say, popular physics

**books**where understanding can only be vaguely metaphorical at best without knowing quite a bit of advanced mathematics giving the illusion of knowledge yes, you ve

**read**The Elegant Universe , but sorry, you still know bupkis about string theory , statistical concepts can be explained and even employed in a critical fashion without much math at all Knowing that variation is much informative than simply the mean doesn t require that you know calculus Likewise for understanding simple experimental design and most experimental designs are simple state a null, apply Student s t test, and you ve got 70% of published scientific papers Of course, saying that something can be explained without math is not the same as actually doing it proficiently, but Wheelan has excelled here The examples are all intuitive, and the writing is clear and easy Perhaps importantly, Wheelan spends an entire chapter on the Central Limit Theorem halfway through the book, and then uses that to explain statistical inference, sampling, and regression Giving the Central Limit Theorem such pride of place is appropriate but is often neglected in basic statistics text

**books**not to mention popular statistics

**books**The book is not flawless, but the quibbles are minor First, Wheelan has a silly sense of humor that intrudes into the book too often culminating in several pointless footnotes that only serve to extend jokes Second, although there are a few mathematical appendices for various chapters, they are generally far too short and actually need math than they have As it is, they are likely to confuse than help.

In general, Wheelan s book is a must

**read**for anyone that hasn t taken a basic stats course so every journalist ever or can t remember much from when they did take it.

A nota foi 3, porque n o poss vel colocar 3.

5 timo livro pra quem est come ando ou apenas se interessa por estat stica Eu estava buscando exatamente um livro como esse para entender melhor conceitos t cnicos O objetivo n o era aprender estat stica a fundo, o que voc n o faria com esse livro, mas entender que estat stica, de verdade, dif cil, n o pela matem tica envolvida, mas porque quem faz estat stica precisa ser atenta, sagaz e sensata O livro vai ficando cada vez mais t cnico da metade pra Frente O in cio explica tudo o que voc pode encontrar na internet como estat stica descritiva, modelos de posi es, etc etc j no final ele apela para as armadilhas das an lises de regress o Algumas anota es Entender o objetivo do que se quer saber com a estat stica Nem sempre os n meros puros conseguem te trazer respostas verdadeiras Voc precisa primeiro entender qual o objetivo que se quiser buscar pra entender se os n meros dispon veis conseguem te dar as respostas esperadas.

Por exemplo descobrir se fumar causa c ncer n o uma tarefa facil N o s separar dois grupos de pessoas que fumam grupo de pesquisa e outra que n o fumam grupo controle e ver no que da Tem que isolar exatamente a diferen a entre os dois grupo O grupo controle pode ter uma dieta ruim ou outros h bitos que podem estragar a pesquisa Coeficiente de Gini mede a o quanto a riqueza ou renda partilhada equitativamente dentro de um pa s numa escala de 0 a 1 A estat stica pode ser calculada para a riquezas ou para a renda anual a n vel individual ou familiar Quanto mais perto de zero, melhor Se toda a riqueza est concentrada em uma fam lia, o coeficiente um e mostra desigualdade Se toda a riqueza est distribu da de forma igual a todos as fam lias do pa s, o coeficiente zero a estat stica descritiva existe para simplificar, o que sempre implica alguma perda de nuance ou detalhe uma fun o chave da estat stica usar os dados que temos sobre perguntas mais amplas para as quais n o temos informa o completa Em suma, podemos usar dados do mundo conhecido para fazer infer ncias informadas sobre o mundo desconhecido estat stica descritiva pode ser totalmente correta, mas pode trazer insights enganosos por causa do erro das m dias o b sico da estat stica encontrar o n mero moderado ou o valor do meio de uma distribui o que pode tentar trazer neutralidade na compara o de dois cen rios A medida mais b sica do meio de uma distribui o a m dia para distribui es sem valores at picos s rios, ou seja, sem valores muito altos ou muito baixos do que os outros valores da distribui o, a mediana e a m dia ser o semelhantes valores absolutos geralmente podem ser interpretados dem qualquer contexto ou informa o adicional Por exemplo saber que algu m precisou dar 80 tacadas para acertar 18 buracos no golfe Com essa informa o absoluta j poss vel avaliar a performance do jogador, sem comparar outro jogador valores relativos do tem significado em compara o com alguma outra coisa, ou num contexto mais amplo Se o golfista est em nono lugar no ranking, podemos compar lo com os oito golfistas que tiverem melhor resultado que ele em um mesmo campo, mesmo hor rio, mesma temperatura, etc Esse resultado comparativo fica muito mais interessante se pegarmos o jogador que precisou das 80 tacadas e compararmos ele numa distribui o de percentil com os outros jogadores, conseguiremos saber se acertar 18 buracos com 80 tacadas algo bom ou ruim Dele cair num percentil de 90, quer dizer que ele est acima de 90% dos jogadores daquele grupo A m dia influ ncia pelas medidas extremas, pelas dispers es Um valor muito alto ou muito baixo pode influenciar a m dia facilmente a mediana n o tem influ ncia de valores extremos, mas ela n o mostra o que est nos extremos do ponto de vista da acusa o, a quest o mediana versus media hora em torno de se os valores extremos numa distribui o distorcem o que est sendo descrito ou s o, ao contrario, parte importante da mensagem se quiser descrever um grupo de n meros de uma forma que o fa a parecer grande, focalize na m dia Se quiser fazer que se pare a menor, focalize a mediana a mediana n o pesa observa es de dist ncia que os n meros se situam do ponto m dio, apenas se est o acima ou abaixo a correla o mede o grau em que dois fen menos est o relacionados entre si a Netflix consegue recomendar filmes para deus usu rios porque ela relaciona os filmes que o usu rio avaliou bem, com filmes assistidos e avalia es e filmes assistidos de outros clientes probabilidade o estudo de eventos e resultados envolvendo um elemento de incerteza probabilidades n o nos dizem o que acontecer com certeza, dizem o que prov vel de acontecer e o que menos provavelmente de acontecer.