November 7, 2012

Math muddle

Sam Smith - One of the reasons we're in such a mess is because most of our politicians and journalists majored in fields that didn't require any competence in math and its relationship to real life. So they tend to treat numbers as adjectives that you can use as freely as, for example, they use the phrase "fiscal cliff." We are about to be inundated with all sorts of statistics aimed at proving why we should slash government services, Social Security and Medicare.

But have you ever heard any public discussion about the difference in spending for capital projects and for operation expenses? Businesses do that all the time, and it could make a huge difference in our politics, but instead they just throw out some scary numbers and you forget that you wouldn't be living in your house without carrying a sizable debt.

Another example: many of those trillions of dollars  you hear about didn't used to exist because there was a time when one talked about an annual budget instead of multiplying by ten years to scare everyone. There's also a certain arrogance in this shift, since Congress and the president only have the right to approve next year's budget.

Then we have the Nate Silver phenomenon in which someone thrills the elite by adding extra decimal pplaces to readily available probabilities. Your editor, for example, was 17 electoral votes short while Silver was ten off. I was also a statistically insignificant 1.23 points different in my popular vote projection.

But what if it had rained all day in Florida? The point is that there is life outside of statistics, which is why I  wouldn't take Silver to Las Vegas with me. A overly precise statistical assumption tends to divert one from other factors, such as the expression on the fact of the guy across the table.

Other thoughts
  • The best way to predict the future is to invent it. - Computer scientist Alan Curtis Kay.
  • I never make predictions, especially about the future - Yogi Berra

Improbable Research - For the third election in a row, the Annals of Improbable Research U.S. Presidential Election Algorithm ( see Debowy and Schulman, AIR Online, 20 October 2003 and Schulman and Debowy, AIR Online, 11 August 2012) correctly predicted the outcome of the United States presidential election. The validity of this algorithm will continue to be tested at a rate of approximately 8 nanohertz for the foreseeable future.

Bonus (somewhat related), from XKCD:

No comments: