Via Tamino at OpenMind:
Suppose you have a child, a son — he’s 10. You want to know whether or not he’s growing normally, so every day you measure his height with a tape measure. You’ve done so since he was 5. You even plot the data on a graph, and notice two things about it. First: the measurements show a fair amount of jitter, sometimes they’re a wee bit higher, sometimes a wee bit lower, there’s noise in the data. Second: there’s also a trend. Your kid is a lot taller at 10 than he was at 5, in fact the trend over the observed time span is upward and reasonably steady. You even do a statistical analysis, estimate the growth rate, and determine that it’s definitely statistically significant — so it’s not a false trend due to noise in the data, it’s real. Your son is growing normally.
Then you’re interviewed by a reporter from the Daily Mail. He asks, “Can you prove — with statistical significance — that your child has been growing since last Tuesday?”
You reply that no, even though the trend over that time span is upward, it’s not statistically significant.
The next day you read the article in the Daily Mail which is titled, “Growthgate U-turn as parent admits: There has been no growth since last Tuesday.”
You protest. “I never said my child wasn’t growing! I just said that the data over such a short time span didn’t show it with statistical significance! That’s only because on such a short time scale, the noise obscures the trend.”
Alas, it’s too late, the damage is done, because 3500 blogs have repeated the article from the Daily Mail and child protective services has been asked to investigate your fitness as a parent.
Gavin Schmidt at RealClimate has this to say about the Daily Mail’s horrible misquoting of Phil Jones:
What Jones actually said is that, while the globe has nominally warmed since 1995, it is difficult to establish the statistical significance of that warming given the short nature of the time interval (1995-present) involved. The warming trend consequently doesn’t quite achieve statistical significance. But it is extremely difficult to establish a statistically significant trend over a time interval as short as 15 years–a point we have made countless times at RealClimate. It is also worth noting that the CRU record indicates slightly less warming than other global temperature estimates such as the GISS record.
It is extremely difficult to establish a statistically significant trend over a time interval as short as 15 years.
As ever, the wonderful Skeptical Science covers the whole affair in more depth.