Writing Issue in Stats Projects

In the new book of student projects, I make the point that students are getting to do technical writing, but that this is not a technical writing course, so I was pretty forgiving. Mostly this was because, for this class (at this time) it would have sucked the life out of the project to insist that students improve the projects after they were turned in and graded. I did give feedback, but not all students (read: few) took the chance to revise the projects for the book.

But if I were ever to do it again, what could I insist on? Where did I think they could improve with better instruction?

Another way to put it is, how would I change the learning goals on which I assess them?

There are a couple of things I noticed about the student work that I want to record here while they’re fresh:

• As adults, there are a lot of things we know that we bring to bear in order to tell what’s reasonable or what’s surprising and interesting. We’re developing that in students, but they’re really just beginning. This is flagrantly true when we look at financial data—they don’t have a clue about what’s a high salary or how much it costs to rent an apartment—but their lack of experience shows up elsewhere too. This is a huge opportunity to turn our classrooms into places where they learn useful stuff.
• As writers, they are not always very skilled in building technical arguments. The reasoning has to be tighter (or at least different) from what they’re being trained to do in history and English. Maybe more to the point, technical writing offers more chances for internal inconsistencies, and those need to be rooted out and expunged.
• As writers, they’re not as willing as they should be to revise. Changing direction seems to be a bigger chore than it should be given that they’re using word processing. Where we might think, no biggie, it’s only five pages, they say, OMG, it’s five whole pages!

A typical example

When we started the projects, a couple of students wanted to document Civil War deaths. (Interestingly, using Census data to make a new estimate of Civil War deaths was just in the news.) To try to do this, one of them picked a southern state and collected data from 1850, 1860, and 1870. He was expecting it to rise from 1850 to 1860 (a kind of baseline increase) and then, if not decline in 1870, at least not increase as much.

Fig. 1. The wrong story

Figure 1 (right) shows what he saw.

He was mightily disappointed; he was hoping that the 1870 bar would go down instead of up. Naturally, I asked what he thought could have caused that, and he said, maybe there were a lot of babies right after the war. Hmm, I said, looks like a lot more people than I would expect to see. If there were a lot of babies, could you see that in the 1870 data?

So we made age distributions for 1860 and 1870, and they looked pretty much the same. So it’s not babies. Keep looking, I told him.

Ultimately, he chose not to focus on population decrease, but rather on how the war affected family structure: he found more widows, more multiple families living together. I applaud this, because part of the point of the project was that frequently, you can’t find the exact data you want, so you need some kind of stand-in: he asked himself what other effect will the war have?

So it was good work, but also a missed opportunity. He did eventually find out about that population increase: to quote the author: “However, it is important to point out that this [1870] was the first time the Negro population was accounted for in the US Census.”

Fig. 2. The story corrected

That is, he could have followed the original graph with Fig. 2 (also at right)—which is exactly the same except the bars are split by race. It shows that the population of whites did decrease (just as he had hoped), and makes plain what we might have thought of before but didn’t: before 1870, we did not even count the slaves.

He noticed this, but didn’t go back and revise. He edited a little, but didn’t change any of the main points of his paragraphs or re-order his argument. And that’s really what was called for. After all, Figure 1 tells a wrong story. And it’s not just a little wrong, it’s grossly wrong. In nonfiction, you shouldn’t say something grossly wrong unless you alert the reader. You could write, for example, “first I thought that the population doubled, but later I found out (Fig. 2) that…”

Again, he did well with what he did, but what would I wish for him next time (if there were a next time):

• Keep Figure 1.
• Explain why there’s no way population grows that much, and how we can tell it’s not babies.
• Put in Figure 2, note that the slaves had not been counted but then calculate a population deficit from the war.
• Follow it up with, for example, splitting that analysis by sex (which seems an obvious next step).

If we adults saw that first graph, we would know that something was fishy. Populations of humans just don’t double in 10 years. Then, when we put race in the graph, we’d know immediately what happened and realize how significant it is as a part of history. This same student picked a Union state for comparison: California. I would never have picked California, though it was a Union state. I know we didn’t send a lot of people to fight (you hear about regiments from Illinois and Massachusetts and Maine. But California? No.) and since gold was discovered in 1849 and statehood was 1851, you know that the population of the state will not be following normal patterns.

Should we expect students to know that? No, but somehow, more feedback would help them find out. And if I did this again, I’d have these papers to use as example, both for what to do (many of them are really ingenious in their use of data) and for places that could be improved.