Practical Statistics for Data Scientists: 50+ Essential Concepts Using R and Python

A very good book, an easy read and covers a lot of basic statistics concepts you would learn in intro to stats university course but in a way more approachable way. If you already good with your stats, you can skip this book. If you feel like stats need improvement, this is a good start.

I had purchased a new physical copy of the book, and realized there were several pages that were blank and missing. I contacted O'Reilly about the problem and they were extremely quick with a resolution! They were able to give me a different copy so I could read it without the missing pages. The content of the book itself is good, except in all black and white, which doesn't bother me personally but may bother someone else when it comes to the graphs. I think the R and Python content are both great, and it keeps the code concise and quick to the point. Great for R beginners, but for python users I would recommend a little more experience. As for the math parts, its great for those who are new to statistics and gives easy to read explanations, and a great refresher for those who just want to review some of the concepts. I especially like the sections provided for further reading, which have been helpful.

I was looking forward to reading this book due to the excellent reviews on Amazon, but it failed to deliver. I struggled to understand the intended target audience, as most of the concepts are normally taught in high school-level courses. In all honesty, the hype surrounding this book speaks volumes about the average knowledge of statistics among Data Scientists.

In my view, this book’s strength is the deep knowledge of the authors added by the ability to explain key points in a few sentences. I love the frequent question and answer to “Is it important for Data Scientists?” Data Science is such a wide and deep topic, that any pointers are extremely welcome. Who is this book for? I believe it’s for intermediate to advanced Data Scientists. There’s so much “wisdom” that any reader should find value in the book. The code snippets are in Python and R. Sometimes those snippets are enough (e.g. power analysis). Sometimes the reader needs different sources to dig deeper (e.g. bootstrapping where I highly recommend infer in R). I believe this “compressed” approach is smart. Data science is too wide and deep and we must be able to dig deeper on our own. In other words, for a beginner, the code is often not enough to learn a new concept. Experienced Data Scientists should be able to judge from the code snippet if it’s enough. +++ Personal highlights: +++ One of the best explanations on effect size I’ve ever seen (page 135). Sometimes, the statistics community uses different terms than the machine learning community. The authors seem to understand both (page 143). For example, in the last 10 years or so, we’ve seen a trend in statistics that favors data and simulations over classical probability theory and complex tests. But why would we use permutations in a hypothesis test? On page 139, the authors explain in succinctly in two sentences. In fact, the authors have a deep knowledge of resampling and how to use simulations over classical tests. The authors don’t try to confuse you. I’ve seen new books which used two pages to explain recall and then two pages to explain sensitivity. In this book, they don’t do it. Recall is the same as sensitivity (page 223). Another example is “Power and Sample Size.” In only four pages, the reader probably gets a good idea of the four moving parts: sample size, effect size, significance level and power. This stuff is hard and explaining it well is even harder. When cluster algorithms tend to give the same results and when not. Funny: “…regression comes with a baggage that is more relevant to its traditional role …”(page 161). Why would a Data Scientist care about heteroskedasticity? Page 183. Kudos

Buen libro con un excelente contenido temático