In a note here last December, I quoted Lord Kelvin’s 1883 admonition that “when you cannot express it in numbers, your knowledge is of a meagre and unsatisfactory kind.” He was right enough that this is still a familiar maxim, 140+ years later – but I find that I already need to hedge my follow-up that “Today, that’s just a starting point, because mere numbers are now the easy part.” I wasn’t completely wrong, but there’s a lot of fuzz on that cue ball – because numbers are definitely easier to get than they were in a world without sensors and connections, but relevant and coherent numbers are a specialized subset that may require considerable effort to refine and understand.

What set me off on this was a rigorous note this week from Andrew Dessler and Zeke Hausfather, writing at The Climate Brink, concerning those who attempt to minimize the severity of climate change by comparing recent temperature events against those of a century ago. Perhaps I should say, “by purporting to compare” – because, Dessler and Hausfather point out, there are numbers that seem to come up often from “Global Historical Climatology Network daily (GHCNd)” records that have a massive structural bias. Showing a map of the locations of measurement stations during the period 1930-1939, D&H observe that

As you can see, the stations are not evenly distributed — there are a lot more stations in the middle of the continent — the exact location where the heat of the 1930s was focused. This is the Dust Bowl region, whose exceptional heat was due to drought exacerbated by poor land management by humans.

Without adjusting for this, the result will be biased towards the hottest part of the continent, thereby overestimating the heat of the 1930s.

By choosing early-decades data from a region of outlier temperatures, the rise of temperature since then is made to seem less. If someone wants to assert some kind of moral superiority by saying, in effect, “at least I’m not manipulating my data,” D&H riposte with

If you do not adjust the data for known biases, you’re getting the wrong answer. [Italics in original] Adjustments and bias correction are not a conspiracy, they’re good science… If you want to claim that the adjustments are wrong/bad science/fraudulent, please be specific about which adjustments should not be made.

I admire the dexterity with which D&H flip the script here: in effect, “I fully disclose that I’ve made the following adjustments; here’s the scientific basis; here’s the resulting algorithm that I used; here’s the code that implements that algorithm; tell me and your audience why this is wrong.”

We’re in a world of decidedly different beliefs and agendas, with abundant access to “alternative facts.” Forevermore, there’s consequently going to be considerable and chaotic argument—whether sincere or specious—over what constitutes a good-faith adjustment of raw numbers, versus what constitutes cherry-picking of the data and misframing of what it means.

Fortuitously, I wasn’t sure if I was using that word “misframing” correctly, so I looked it up: I found that its first use is quite definitely traced to Sir Thomas More’s 1533 essay “Apology,” where he wrote (paraphrasing slightly for brevity and clarity, and adding my own boldface):

If a writer proveth a point by a ‘some say,’ he showeth himself not indifferent, when he bringeth in the one and leaveth the other out… misframing his matter more toward division than unity…

Note that More wrote this 350 years before William Thomson (Lord Kelvin) got around to talking about putting one’s argument into numbers, which leads me to feel that Thomson missed an opportunity to be even more quotable today. Imagine the good that might have been done if we were all to cite him as having said “when you cannot express it in numbers, your knowledge is meagre; when you do not express it in accurate and clarifying numbers, your expression is unsatisfactory.”