Results tagged “statistics”

It's being reported all over that the unemployment rate in Washington "dipped" by 0.1 percent, from 9.2 to 9.1 percent. This month-to-month non-change is far more accurately labeled stasis than a "dip"--the real news is that of the 315,000 or so unemployed, thousands are one month closer to running out of UI benefits. Unemployment figures have been politically distorted to the point of absurdity--for one thing, people who fail to find a job for so long that they give up are no longer counted as unemployed. The precision of 0.1 percent change would be laughable if it weren't for the cruel irony that making these announcements is a real job.

The stats geeks at Deconstructing Sundance aim to be the FiveThirtyEight of the film festival world by predicting the future box office success of Sundance films, using the words in their festival guide descriptions (and a Bayesian algorithm) alone. Example: 62% of Sundance competition films in the last 15 years whose descriptions included the words "gay," "gays," or "homosexual" went on to commercial success. So "gay" becomes a positive indicator of success, and so on and so forth.

When SPD reports came out earlier this month with the news that major crime in Seattle has gone down by some 11%, the Southeast Seattle Crime Prevention Council was skeptical. They crunched the numbers on their own neighborhood themselves, and found some startling results: assaults, homicides, and non-residential burglaries have all increased by huge numbers. Though even the police chief went on record this week to speak on Seattle's deepening gang violence problems, the police department is refuting the SSCPC's numbers, saying they aren't taking all the factors into account (precinct boundary changes and incidence of "actual crime" versus dispatch calls).