Set up your notebook. You will need it at the end of class.
Copy these. Leave space under each one for a definition.
- bucket
- univariate
- claim
Think about these. Nothing to write.
- When you tell someone how often you do something, do you round up or down?
- Is it easier to guess about yourself, or about other people?
Copy this into your Lab Notebook and leave room under it. You will answer it at the end of class.
Why can two survey items about the same thing disagree?
A person holding a stack of survey results faces a problem before they face any question about what the results mean. The stack is not organized. Some items in it speak directly to whatever the person wants to know. Some speak to it sideways. Some do not speak to it at all, even though they concern the same general subject. Before any of it can be used, it has to be sorted, and how it gets sorted determines what conclusions are available afterward.
The sorting categories are called buckets, and choosing them is the real work. A bucket is not a topic. It is a kind of evidence. Two items can both be about school attendance and still belong in different buckets, because one asks a student how often they come to school and the other asks a teacher how often that student comes to school. Same subject, different kind of evidence, different bucket.
This matters because the buckets are what make disagreement visible. If every item goes into one undifferentiated pile labeled "attendance," and the pile contains both student answers and teacher answers, then a researcher who summarizes the pile will produce an average that hides the fact that the two groups said different things. Sorted into separate buckets, the disagreement becomes the finding rather than being smoothed away.
Every good bucket system needs one bucket that most people forget to build. Some items in the stack will not answer the question at all. They are related, they are interesting, and they are irrelevant to the specific thing being asked. An item recording which brand of shoes students wear is genuinely about students and genuinely useless for a question about attendance. That item needs somewhere to go, and the somewhere cannot be a bucket that implies it counts as evidence. Researchers who do not build a reject bucket end up quietly stretching a good bucket to hold something that does not belong in it, and the stretch is invisible in the final report.
Deciding an item does not answer the question is a judgment, and it is a judgment that has to be defended. The defense usually comes down to a difference between what the item asked and what the question asks. A question about how often something happens is not answered by an item about whether it happens at all. A question about how much is not answered by an item about which. Naming that mismatch out loud is the difference between rejecting an item and dismissing it.
So far, everything described here counts one item at a time. A researcher looks at an item, reads the counts under each answer, and decides what that item shows. This is called univariate work, from a root meaning one variable. Almost all reporting on surveys is univariate. Sixty-two percent said this. Forty said that. Each number describes the answers to a single question.
Univariate work is not a lesser form of analysis. It is the foundation, and most published survey findings never go beyond it. But it has a specific limit that is worth naming early. A univariate count tells you how many people gave each answer. It cannot tell you anything about which people. Two items counted separately show two distributions and nothing about the relationship between them, because counting one item at a time discards the information about who said what.
Once items are sorted and counted, the last step is to say something. That statement is a claim, and a claim is not the same thing as a guess or an opinion. A claim is a statement that specific evidence supports and that could be shown wrong by other evidence. Both halves matter. A statement nothing supports is a guess. A statement nothing could contradict is not a claim either, because a statement that survives every possible piece of evidence is not saying anything about the world.
A well-built claim carries its evidence with it. Instead of asserting that attendance is a problem, it names which items support that and what those items actually asked. Doing so makes the claim checkable, which is the entire point. A reader who disagrees can go look at the same items and argue about them. A reader facing an unsupported assertion can only agree or refuse.
The order is fixed and it does not reverse. Sort first, count second, claim third. A researcher who forms the claim first will sort the evidence toward it without noticing, and will build the buckets that produce the answer they already have. This is not usually dishonesty. It is what happens when the conclusion arrives before the categories do.
B. A bucket is a kind of evidence, not a topic. Same subject, different source, different bucket. This is the whole reason our three buckets are About me, About others, and Doesn’t answer instead of being named after topics.
C. The item has to go somewhere, so it goes somewhere wrong. Nobody reading the final report can see it happened. Our Doesn’t answer bucket exists so that never has to happen.
B. Counting one item at a time throws away who said what. The passage says two items counted separately show two distributions and nothing about the relationship between them. Hold onto this. It is what tomorrow is about.
B. A statement that survives every possible piece of evidence is not describing the world. The passage puts the two halves side by side: a statement nothing supports is a guess, and a statement nothing could contradict is not a claim either.
B. Sorted into separate buckets, the disagreement becomes the finding. In one pile it gets smoothed away, and nothing in the final number shows that it ever happened.
C. And the passage draws the line precisely: naming that mismatch out loud is the difference between rejecting an item and dismissing it.
A. The passage gives two of this shape. A question about how often something happens is not answered by an item about whether it happens at all, and a question about how much is not answered by an item about which.
B. And the passage says why that matters. It makes the claim checkable: a reader who disagrees can go look at the same items and argue about them, while a reader facing an unsupported assertion can only agree or refuse.
B. They build the buckets that produce the answer they already have. The passage’s phrase for the result is that the conclusion arrived before the categories did.
B. The passage says plainly that this is not usually dishonesty. That is what makes the fixed order worth having as a procedure rather than a warning about character: it protects you from something you would not catch yourself doing.
- Never10
- Once or twice44
- About weekly12
- Almost daily12
- Skipped1
- 04
- 10
- 22
- 31
- 41
- 56
- 61
- 76
- 89
- 98
- 1040
- Skipped1
- Not at all common4
- Not too common6
- Somewhat common20
- Very common48
- Skipped1
- Much less24
- Somewhat less40
- About the same12
- Somewhat more or much more2
- Skipped1
- ChatGPT70
- Gemini36
- Grok7
- Copilot6
- None of these4
- Claude2
- Perplexity1
- DeepSeek1
A06 asks about you. And look at the spread — 10, 44, 12, 12. High variability.
A17 asks you to estimate everybody else. That is a guess about a group, not a statement about yourself.
It is about AI, and it does not tell us how much this class uses it. Knowing which tool someone opened says nothing about how often.
A01 is also the only item on the survey where you could pick more than one, which is why its counts do not add to 79.
A17: 40 of 79 say all ten of their classmates use it without permission.
What do you make of that?
We are not answering this today. Write down what you think and bring it back tomorrow.
Every item so far, we counted one at a time. A06 alone. A17 alone. A20 alone.
That is what the passage called univariate. It tells you how many people gave each answer, and nothing about which people.
Partner work first. Then quiet work time.
- No cellphones
- No sleeping
- No non-academic use of Chromebooks
- A quiet room and music
- No interruptions from the teacher
- If you are caught up on this course's work, you may work on anything legitimate for another class