Data Dive · Rung 3 · What's Related · 3.6

One measure is not a profile

Two adults describe the same students and disagree about who is struggling. Both of them are right about their own measure.

Runs in 30 minutes · the chart needs nothing · students open the pool for the middle third, a browser each and no accounts

3.6→ next · ← back · esc leaves0 / 5

1 - The scene

Nothing on the screen yet

The counsellor says the students struggling most are the ones getting bullied.

The nurse says the students struggling most are the ones who never get to play a sport.

They are describing the same students, in the same building, and they cannot both be looking at the right measure.

Take a show of hands, and write the split on the boardCounsellor or nurse? Do it before anything is drawn. A chart shown first is a chart to be admired; a chart shown second is a chart to be used.
3.6 · One measure is not a profile→ next · ← back · esc leaves1 / 5

2 - The look

Twenty-nine groups, ten measures, one line each

Twenty-nine groups of American teenagers, ten measures eachCDC National Health Interview Survey - one line is a group, not a person▲ worsebullied55.925.4▲ worseanxiety43.312.3▲ worsedepression37.310.1▼ worseregularbedtime74.550.4▼ worseactive 60min69.849▼ worsesportsteam7439▼ worseparentsupport73.141.6▼ worsepeersupport56.743.3▲ worsescreentime2h98.185.7▲ worsefair orpoorhealth14.65.5the other 26 groupsthe two richer income groupsteens under 200% of the poverty line

Each line crosses every measure at once. The three family-income groups are picked out; the poorest is in yellow.

29.8%
poorest teens bullied
34.6%
richest teens bullied
-4.8
the gap, pointing the wrong way

Say what a line is, once, slowly: this single line is every teenager in the country under 200% of the poverty line, and it has a height on all ten axes.

Every axis is scaled to its own range, and they do not all point the same way — the small caret above each name says which end is the bad one.

Do the subtraction on the board, then stopThe poorest teenagers report the least bullying of the three income groups. On that one measure they are the group doing best. Let it sit there before anybody opens anything.
3.6 · One measure is not a profile→ next · ← back · esc leaves2 / 5

3 - The work

In the pool, on the same file

Pairs, twelve to fifteen minutes. Every answer is findable, and the last one has more than one right answer.

Open the pool on this file and do all five
  1. Hover the line for Less than 200% FPL to isolate it, and follow it left to right.
  2. Write down which way each axis points before you count anything. Five of the ten are bad when they are high.
  3. Count. On how many of the ten measures is the poorest group the worst of the three income groups?
  4. Drag the axes so bullied sits next to sports team, and watch what the yellow line does between them.
  5. Find one other group whose line crosses the same way — better than the field on one measure, worse on most of the rest.

Step four is the one to watch for. Between those two axes the yellow line falls from the top of the field to the bottom of it, and that fall is the finding: no chart of a single measure can hold it.

3.6 · One measure is not a profile→ next · ← back · esc leaves3 / 5

4 - The say

One sentence each, out loud

8 of 10
measures the poorest group is worst on
2
it looks best on: bullied and screen time
28.5
points behind on sports team

I think <which adult is closer> because <what is on the chart> and I checked it by <what I did to the chart>.

The third clause is the one that counts. “Because the lines cross” is somebody who watched. “Because I put sports team next to bullied and the line that was lowest on one was highest on the other” is somebody who worked — and it is a sentence no machine can write for them, because it names what they did to their own screen.

One measure is not a profile. A group can be genuinely best on one axis and genuinely worst on eight, and anybody who reports only the first axis has told you the truth and misled you completely.

Thirty seconds at the end, and do not answer itBullying here is self-reported. A lower number can mean less bullying, or it can mean fewer children naming it as bullying. Ask which, and leave it open.
The three questions
  1. Which measure would you have chosen if you only had room for one, and what would it have hidden?
  2. Name the two measures the poorest group looks best on. Are they good news?
  3. Somebody hands you a chart of one number about a group of people. What is the first thing you now ask them for?
3.6 · One measure is not a profile→ next · ← back · esc leaves4 / 5

After the dive

The shapes on this page, in the library’s own words

parallel coordinates

Each line is one thing crossing every measure. The line that cuts against the others is the odd one.

What it leaves outOnly neighbouring axes can really be compared, so the order you put them in decides which relationships a reader can see.

library · ladder · dives · the pool · this dive’s worksheet

3.6→ next · ← back · esc leaves5 / 5