Nate Silver writes about the power of statistics

I have a weird passion—I love the subject of statistics.  When I was at the University of Minnesota, it was possible to get a BS degree from the College of Liberal Arts (instead of a BA) if you took the stats sequence which included three quarters at the upperclass level and one quarter at the graduate level.  UM was VERY proud of its role in introducing statistics to the social sciences.  The psyche department had developed a wildly profitable test called the Minnesota Multiphasic Personality Inventory in 1943 that used regression analysis.  In those days, the math was done on slide rules.  Because the military found the test useful, it became so widely used, it is a rare American who hasn't taken the test at least once.  The UM collects royalties to this day.

So here's the scene I stumbled into.  The social science departments at UM believed in their bones that stats were the key to fame and fortune.  On the other side of the campus, the computer science department had become a feeder system for the area manufacturers such as Honeywell, Control Data, and Cray so were also full of themselves.  The social science guys hated doing regression analysis on slide rules.  Someone had written a piece of software called the Statistical Package for the Social Sciences that did the necessary computations in seconds rather than months.  So the Social Sciences had managed to get their hands on an IBM 360, the computer jockeys got everything working, and someone like me got to play with a world-class computer for a $5 / quarter lab fee.  I mean, what wasn't to love?

The problem with this academic circle-jerk was that a whole bunch of people came to believe that you could generate questions almost at random and if you asked them to a well-selected sample, math would discover the meaningful questions and math would discover the important correlations between them.  You could look at a question with no historical background, no experience, and no expertise and still find important answers because, by gum, you had an IBM 360 by your side.

While stats are so powerful they literally form a new branch of epistemology, they are far more likely to be used to confuse and deceive. But even when stats are used by people who are genuinely trying to understand their world, there are just folks who do it better than others.  And probably the most elite practitioner in the world of stats these days is Nate Silver—who has written a new book on how to tell the difference between important data and the noise.  When Silver argues that "that we frequently go wrong in many areas when we adopt a single model or approach to a problem, when an evolving, flexible, multiple-input, probabilistic approach would serve us better" it comes as music to me because this is why I argue that heterodox economics is always superior to orthodox economics and that the most important habit of thought is evolution or continuous improvement.

Book Review: The Signal And The Noise by Nate Silver

By The Banker 3 Oct 2013

I took a mandatory course in high school[1] called “Theory of Knowledge,” meant to help us consider ‘How do we know things?”

“How do we know things?” turns out to be one of those big philosophical questions – dating from the time of Plato & Aristotle – irritating all us for the last few millenia.

What Nate Silver addresses more than anything in The Signal and The Noise: Why So Many Prediction Fail – But Some Don’t is how we know things – in particular how we use and misuse information to understand and make predictions about complex phenomena such as baseball performance, political outcomes, the weather, earthquakes, terrorist attacks, chess, Texas Hold ‘em poker, climate change, the spread of infectious diseases, and financial markets.

I’ve written before that it’s Nate Silver’s world, and we just live in it.[2] The Signal and The Noise offers a 21st Century answer to the question of ‘how do we know things?’ Because most of us, and most media, do not yet think this way, Silver implicitly criticizes everything I hate about the Financial Infotainment Industrial Complex.

Big Ideas vs. Small Ideas

Silver argues effectively that we frequently go wrong in many areas when we adopt a single model or approach to a problem, when an evolving, flexible, multiple-input, probabilistic approach would serve us better.

The problem of political pundits

Silver repeatedly returns in The Signal and the Noise to criticize political pundits on a TV show called The McLaughlin Group, on which commentators from the left and the right appear to make bold political predictions. Silver – among the most widely admired public forecasters of political outcomes – eviscerates this type of ‘prediction,’ citing data that shows these commentators make accurate predictions no more often than would a random coin toss.

But television rewards ‘bold stances’ and ‘big ideas’ of the type The McLaughlin Group traffics in, while largely ignoring more thoughtful approaches.

Silver labels and criticizes the “Big Idea” mindset that passes for political commentary on television in favor of a more modest, probabilistic, and empirical “Small Idea” mindset. Small ideas, nuanced, uncertain, and modest, however, make for poor television ratings.

But Silver does have a Big Idea himself

For complex, hard to predict phenomena[3], Silver explains his preferred method, based on a probability theorem attributed to an 18th Century English minister Thomas Bayes.

No doubt Silver thinks many more of us should become familiar with this branch of probability and statistics math. [4]

Beyond the Bayesian theory, however, Silver encourages us to adopt a probabilistic world-view. His big idea is for us to move away from “I have the explanation and I know what’s going to happen,” to a different way of understanding the world characterized by “I can articulate a range of outcomes and attach meaningful probabilities to the possible outcomes.”

Over time, as we refine our data gathering and multifaceted models, we can move our small ideas forward and become ‘less wrong’ about the world.  more