Skip to 0 minutes and 13 seconds OK, I’m here with Nicholas Jewell, who is a professor of biostatistics at Berkeley, and soon actually at the London School of Tropical Medicine. And we want to take advantage of this statistical expertise to talk a little bit about what role statistics can have in accounting for death in war. Let me just ask you, what do you think that statistics as a field, speaking very generally, has to contribute to accounting for death in war? Well, I’m interested in what you said about your introduction in the sense that the basic raw data, the documentation you just described, of course, is absolutely critical to anything that follows.
Skip to 1 minute and 6 seconds And I think it’s a little bit of a myth that somehow statistics can clean up messy data, and make up for basic flaws in the way the data is documented and collected. So right there from the start, I think statistics has a role in informing how you do that documentation well, how you de-duplicate records, whether you got multiple copies of the same information. Statistics has a role there in actually getting high-quality data, let alone moving to, as you described, the estimation phase of a project. I really appreciate that point. So that’s a really good point. Perhaps not universally shared among statisticians, I think maybe some would take the position that they come after the data is collected.
Skip to 1 minute and 59 seconds I don’t know how you react to that. Yes. Well, Fisher, the great statistician of the 20th century, always used to say that a lot of his job was performing the autopsy for telling you why the data collection died. [LAUGHS] [LAUGHS] Rather than actually bringing anything to life. So showing up at the end and saying, you boneheads. Why did you do it that way? Exactly. We don’t want to be doing that. Obviously, we would like to involve the statistician early so that you can collect things in the right ways. So understanding war or casualty estimation techniques seems to me to start right at the beginning.
Skip to 2 minutes and 35 seconds So people who understand what you’re teaching in this course would be informed in the future about good techniques, good methods, to ensure high-quality data because it’s not a question of just beachcombing for information, and somehow statistics waves a wand, and produces high-quality information. Poor-quality documentation will lead to very poor-quality estimates in general. And there’s no escape from that. If there’s a fundamental bias baked into the data, the most statistics can do is inform you of that and suggest how far off you might be because of the bias. But it’s much, much better to get those biases controlled and removed in the design of how you collect data from the very beginning. So that’s the first thing.
Skip to 3 minutes and 28 seconds And I think the second thing that statistics brings to the table in this, other than a little bit of mathematics and algebra to show how you manipulate numbers to produce something that the raw documentation itself doesn’t immediately provide, but over and above that, statistics provides you a template for how you’re going to express your uncertainty. And I think it’s really important that uncertainty comes along with any estimate because to some extent, many people just want a number at the end of the day. They want to know how many people were in the crowd. They want to know how many people died in this particular conflict. They may want to know characteristics of that.
Skip to 4 minutes and 19 seconds But all of those numbers really come necessarily with uncertainty. And quantifying that uncertainty is really important. How accurate really is this? Because the same number might be extremely precise, and then you can really debate what it’s telling you. But the same number, if you knew it was very imprecise, you really shouldn’t try and debate what it’s telling you, because you’re saying, we don’t really know. But what is even more important to me as history begins to layer over these estimates and numbers as time passes, the uncertainty gets washed away far quicker than the actual number. And so there’s a famous poem that ends with the sentence, “History counts its skeletons with round numbers.”
Skip to 5 minutes and 9 seconds [LAUGHS] Which is reflecting that there’s a tendency to say, oh, there were a million people killed. Well of course, everyone knows there wasn’t really a million, but maybe when those numbers were first created, there was some range of uncertainty. But that, as I said, tends to disappear in history. And all that remains is the single number. And that’s really unfortunate.
Nicholas Jewell on the connection between documentation and estimation
In this clip I interview Nicholas Jewell, long of the University of California Berkeley and now at the London School of Hygiene and Tropical Medicine.
A notable feature of this interview is the stress Nicholas places on his view that statisticians cannot miraculously overcome the deficiencies of bad data. Thus, he emphasizes the importance of solid documentation of war deaths, thereby creating a strong link between Week 1 of our course and Week 2.
Here an extract from the poem by Wislawa Szymborska that Nicholas cites toward the end of the interview.
History counts its skeletons in round numbers.
A thousand and one remains a thousand,
as though the one had never existed:
an imaginary embryo, an empty cradle,
an ABC never read,
air that laughs, cries, grows,
emptiness running down steps toward the garden,
nobody’s place in the line
Perhaps you would like to discuss this extract in the comments section below.
© Royal Holloway, University of London