The Model Thinker

I must admit that when I first started looking through The Model Thinker, I was hopeful. Here was a book that attempted to explain the use of a variety of different models in order to better understand a real world problem. All the basics and buzzwords are included, from basic statistics through reinforcement learning. Although not aimed at mathematicians, it surely could be useful to business people seeking to understand what this is all about, or to students looking to supplement their courses with material about what no single program has time to cover.

But then…

In a section about linear models (regression):

Significance (p-value): The probability that the sign on the coefficient is nonzero.

That certainly tempered my optimism.

Then “in 2016 the average wealth of white families (approximately $110,000) was more than ten times that of African Americans and Latino families.” I followed the reference but it pointed to something written in 2013.

Finally, although the author starts the book by extolling the value of using many models, there are no examples of using a variety of models to analyze various aspects of a single problem and showing the value that this approach brings. Rather it is a catalogue of many different model types.

As it stands I can not recommend this book.

Peter Rabinovitch is the head of Data Science at ZetaTango Technology, and has been doing data science since long before “data science” was a thing.