Lies, damned lies. . .

 作者:宗糸     |      日期:2019-03-07 06:20:09
By Charles Seife METEOROLOGISTS searching for weekly patterns in our weather may be deluding themselves. After poring over 45 years’ worth of temperature measurements, a statistician in Colorado has found that Saturdays and Sundays tend to be the warmest days of the week. But if he arbitrarily chose Wednesday as the last day of the week— rather than Saturday—Wednesdays and Thursdays emerged as the balmiest. This apparent paradox might help explain odd weekly patterns that have emerged from statistical analyses of the weather, such as why it seems to rain more on weekends (New Scientist, 8 August 1998, p 22). Kevin Coakley of the National Institute of Standards and Technology in Boulder analysed temperature readings from San Francisco International Airport. He found that the quirk depends on the fact that daily temperatures don’t vary randomly—but tend to rise or fall steadily over time. Say the temperature is rising steadily over an extended period of time. If you pick any block of seven days, the last day is going to be the warmest—no matter whether your block begins on a Sunday or a Wednesday or any other day. Likewise, if the temperature is dropping steadily, the warmest day of any block will be the first day. Even if there’s no long-term warming or cooling trend, daily temperatures often generally rise or fall over periods longer than a week. “The values on consecutive days often follow a trend,” says Coakley. “Half the time it’s upward, half the time downward.” And that means the highest temperature is likely to occur on the first or the last day of the week—regardless of its starting point. Having an arbitrary seven-day week may therefore be a problem in analysing weather trends. “It’s definitely something we want to look at,” says climatologist Randall Cerveny of Arizona State University in Tempe,