Not long ago, the Hadley Centre/Climate Research Unit in the U.K. released their new “HadCRUT4″ global temperature data set. That prompted David Rose of the Daily Mail to claim that “Global warming stopped 16 years ago.”
Here, according to Rose, is “the chart to prove it”:
There are some minor technical problems here. For one thing, the graph doesn’t show the last 16 years, it shows the last 15 years. For another thing, it isn’t “tenths of a degree above and below 14C world average,” it’s degrees Celsius above the 1961-1990 world average. Also, even though the first point is labelled “1997″ and the last “2012,” the axis is a bit off because the first point is September 1997 and the last is August 2012.
But those are minor technical problems.
The major problem is that David Rose has made one of the most common mistakes studying data. He looked at a graph and concluded that the long-term temperature trend had changed around mid-1997, then showed only the data since mid-1997 and claimed it was “the chart that proves it.”
His choice to start with mid-1997 was made because that gave him the result he wanted. That’s a practice called “cherry-picking.”
The U.K. Met office replied that his choice to start with mid-1997 was cherry-picking.
David Rose has answered
Q Did The Mail on Sunday ‘cherry-pick’ data to disguise an underlying warming trend?
A Some critics claim this newspaper misled readers by choosing start and end dates that hide the continued warming.
In fact, we looked at the period since 1997 because that’s when the previous warming trend stopped, and our graph ended in August 2012 because that is the last month for which Hadcrut 4 figures were available.
Apparently David Rose isn’t aware of this, but he has actually admitted that yes, the choice was cherry-picking — it was made specifically because of the result it gave.
If he could prove, statistically, that the previous warming trend stopped (or even slowed) around that time, then choosing mid-1997 to start an analysis would be legitimate. But he can’t. Because it didn’t. And then of course there’s the pesky fact that he didn’t even try.
Let’s put the last 15 years (of HadCRUT4 data) into context. Here’s the data since 1975:
I chose 1975 because you can prove (statistically, and there’s no doubt) that the trend changed at that time.
Notice that in addition to its long-term trend, there are also a lot of wiggles up and down. If you want to claim “no warming” whether it’s true or not, then you’d only show part of the data and you’d start when one of those wiggles was well above the overall trend line. Here’s where David Rose chose to start:
That’s not just cherry-picking, it’s championship cherry-picking.
Did the previous warming trend change then? One way to investigate is to “do the math.” David Rose didn’t. I did. It turns out that the trend using just the data since mid-1997 does not disagree (statistically) with the trend prior to mid-1997. You do have to take autocorrelation into account, so it’s not the simplest procedure. But we’ve learned how to handle that. It’s called “statistics.”
Perhaps the most interesting part is that even if the “bare” trend since mid-1997 did disagree (statistically) with the trend prior to mid-1997, it still wouldn’t prove it. That’s because mid-1997 was chosen because of the result it gives. If you just analyze some data (all of it), you apply standard statistical procedures. But if you’re allowed to try a lot of different starting times then choose the one you want (if you’re allowed to cherry-pick), it alters the statistical behavior. Dramatically. Those who are interested in the details should read this. But in this case, the result is so clear-cut we don’t even have to take the cherry-picking correction into account.
We can even look directly at how temperature since mid-1997 compares to the previous warming trend. Here’s the previous warming trend estimated by linear regression:
What would have happened if that trend continued unabated? This:
What would have happened is pretty much what did happen. Imagine that.
Of course global temperature didn’t follow the trend line exactly. It never does. It shows those inexorable wiggles up and down. It so happens that the last couple of years have shown us one of the “down” wiggles. But it’s no more “down” than the 1997-1998 wiggle is “up.” They’re both well within what would be expected, given the degree of fluctuation which global temperature shows all the time.
David Rose is taking advantage of the fact that we’re in a “down” wiggle right now, exploiting the fact that most people (like readers of the Daily Mail) don’t really understand that fluctuations like that are entirely natural, in fact unavoidable, and happen all the time. You can’t take them as evidence of a change in the trend — unless you’re ignorant of statistics and you’re pushing an agenda.
David Rose is also taking advantage of the extreme “up” wiggle in 1997-1998. It too is entirely natural, and is not evidence of a change in the trend — or perhaps David Rose would like to publish a story with the headline “Global warming took off like a bat out of hell in 1997!”
What’s worse, the choice to start with mid-1997 was made specifically because of the result it gives. David Rose himself, said so explicitly. That’s called cherry-picking.
It’s entirely believable that someone could make this mistake quite naturally, for no other reason than looking at a graph and drawing the wrong conclusion. People do it all the time, that’s why we’ve had to work so hard for centuries to develop the science of statistics.
But in David Rose’s case, I don’t think that’s so. My opinion: David Rose came to that conclusion because he has a political agenda and he’s constantly showing “confirmation bias” — he’ll go out of his way to find something he thinks is evidence to support his agenda (whether it really is or not). Since one of the main tactics for this practice is cherry-picking, that’s one of David Rose’s main methods. And he has just shown us a textbook case.
Just my opinion.