top of page

Prediction Revisited also:

​

  • Illuminates many commonly accepted but less commonly understood notions of statistics

​

  • Reveals several valuable yet previously unrecognized mathematical equivalences

​

  • Includes colorful biographical sketches of some of the key scientists whose contributions paved the path to relevance-based prediction

​

  • Enables access to the mathematically minded reader as well as those who prefer an intuitive and conceptual discussion of the book's key ideas

Prediction Revisited: The Importance of Observation

​

Classical statistics originated more than two centuries ago to facilitate navigation by predicting the motion of heavenly bodies and to offer guidance for games of chance. These systems obey relatively simple rules. Today, we are challenged to predict outcomes that are driven by far more complex systems, such as the dynamics of social behavior. Researchers have come to recognize that classical statistics cannot accommodate the complexity of social dynamics; they have therefore turned to the emergent field of machine learning. But they also struggle with machine learning algorithms because these algorithms are often opaque and unintuitive, and they lack a theoretical core. This book offers another way forward — a way that is theoretically grounded, transparent, and intuitive.

​

This path forward requires a new perspective. We must view data as experiences and think of variables as attributes for describing those experiences. And we must recognize that some experiences are more relevant than others. Indeed, determining relevance is the essence of prediction. The authors provide a guided tour of this groundbreaking insight, from its foundations in information theory to its central role in forecasting. They reveal the specific components of relevance and show how to measure them, not only in concept but with mathematical precision. 

​

There is a practical reward to this journey. You will learn that the prediction from a linear regression equation is equivalent to a relevance-weighted average of past outcomes. This critical insight enables you to form more reliable predictions from a subset of the most relevant observations, using an approach called partial sample regression. And you will learn how to judge the unique reliability of an individual prediction separately from the reliability of a prediction model. This more nuanced measure of reliability will enable you to fine tune how you respond to a specific prediction. Aside from these practical benefits, this book will please the intellectually curious by illuminating commonly accepted but less commonly understood notions of statistics and by revealing several obscure but profound mathematical equivalences. 

​

This book is written for anyone who relies on data to contemplate the future, including social scientists, medical researchers, and sports analytics professionals. And in deference to both those who think intuitively and those who think mathematically, each chapter is presented twice: once as a conceptual narrative and again in the more precise language of mathematics. As the reader you are free to choose whichever path suits you, or you can travel along both. Either way, you will discover a new and powerful approach to prediction that is rigorous yet intuitive. 

​

 

© 2024 by Megan Czasonis, Mark Kritzman, and David Turkington

bottom of page