Taking care in writing with statistics

This morning's Wall Street Journal has an article on page A3 entitled "Computer-Based Tests Put Typing Back on the Curriculum" (on the web, it is "Common Core-Linked Tests Spur Schools to Teach Typing" — subscription required), an interesting discussion of how the computer-based testing associated with the Common Core curriculum implicitly assumes sufficiently good skills in typing to write answers for the tests.

One of the challenges described is the disparity in school support for computing, particularly between higher and lower income areas. The article states:

The most recent numbers from the U.S. Education Department show that schools with 75% of students eligible for free or reduced-price lunch have, on average, 170 computers in the school, compared with 209 computers in schools with less than 35% of students eligible.

The data suggest a problem, and the conclusion is probably correct. Unfortunately, the data as presented do not give sufficient information for us to reach that conclusion: we don't know how the schools compare in size. The most useful measure here would probably be the ratio of students to computers. A school with 1000 students and 170 computers is probably in better shape than a school with 2000 students and 209 computers.

Perhaps this was a problem in the Department of Education's presentation of the data; we don't have enough information about it to know.

In any case, when we write with statistics and other data, it's important to make sure that the story from the data supports the points we are trying to make.