Why the Kind of Question Matters
A topic does not determine whether a question is statistical. “Study habits,” “symptoms,” and “campus facilities” could each lead to a question with one fixed factual answer or to a question whose answer needs to account for differences among individuals. The key is not whether the topic is important, or even whether someone needs to look up information. The key is whether the question anticipates variation in the data used to answer it.
This distinction helps clarify what kind of answer a question calls for. A fixed-answer question can be answered with one fact, such as a number, a name, or a yes-or-no response. A statistical question calls for information about responses that may differ across individuals. Its answer must describe or take account of those differences, rather than treating one observation as sufficient.
“Anticipates variability” does not mean the responses must turn out to be different. It means that, before seeing the data, the question allows for the possibility that they could differ and asks for an answer that accounts for that variation. Conversely, a question does not become statistical just because the answer is unknown, difficult to find, or obtained by consulting several records.
A Practical Test
To classify a question, identify what would count as an answer. Would one fact settle it, or would answering it require considering responses from multiple individuals and how those responses may differ? If the latter, the question is statistical. If one fact settles it, it is fixed-answer.
This test is about the question, not the outcome of the data collection. A statistical question can happen to produce identical answers for everyone observed. And a fixed-answer question can take considerable effort to answer: checking several records may be necessary to verify a single total, for example, without asking about variation across individuals.
Ask whether the question seeks one fact or a description that must account for different responses.
For a question about responses across individuals, ask whether those responses could differ in a way relevant to the answer.
If the answer must account for possible variation in the data, it is statistical. If one factual answer settles it, it is fixed-answer.
Worked Examples
Worked Example: Study Habits
Consider two questions about studying: (1) “How many days did each of six learners study last week?” and (2) “How many study sessions are listed on the course schedule?” The numbers below are hypothetical.
For the first question, suppose the six learners report 2, 5, 3, 0, 4, and 2 study days. The responses are not all the same: for instance, one learner reports 0 days and another reports 5. More importantly, the question asks for responses from several learners and calls for an answer that accounts for the possible differences between them. It is statistical.
For the second question, suppose the course schedule lists 12 study sessions. The question asks for one count about the schedule, not for how study behavior varies among learners. The answer is 12 sessions, so it is fixed-answer.
The presence of a number in both answers does not decide the classification. Nor does the fact that the first question concerns last week and the second concerns a schedule. What matters is that one question asks about potentially varying responses across learners, while the other asks for one fact about a schedule.
Worked Example: Symptoms
Compare “Did the intake form ask about nausea?” with “Did patients report nausea?” The following hypothetical responses are for eight patients: Yes, No, No, Yes, No, No, Yes, No.
The question about the form has one factual answer. If the form includes a question about nausea, the answer is Yes; otherwise, it is No. It does not ask how responses differ across patients, so it is fixed-answer.
The question about patients anticipates that different patients might answer differently. In the hypothetical data, three of the eight report nausea and five do not. Because the answer must account for variation across patients rather than give one response for a single patient, this is a statistical question.
The actual responses happen to show variation: there are both Yes and No answers. But this result is not what makes the question statistical. The question was statistical before the responses were collected, because different patients could have different answers and the question asks about their responses.
Worked Example: Campus Facilities
Compare “How many study rooms are listed in the campus directory?” with “How long do visitors wait to use a study room?” Suppose a hypothetical directory lists 14 rooms. Suppose five hypothetical visitors report waits of 0, 6, 3, 10, and 1 minute.
The directory question asks for one count. Checking the directory may take time, but its answer is a single factual total: 14 rooms. It is fixed-answer.
The visitor question asks about wait times across individuals. The five hypothetical values range from 0 to 10 minutes, so they illustrate possible differences among visitors. The question is statistical because its answer needs to account for those potentially varying experiences.
It would not be enough to answer the visitor question by giving one visitor’s wait, such as 6 minutes, as though it settled the question for everyone. One observation does not account for possible variation among the visitors being considered.
Why the Distinction Works
The classification can be justified by considering what information an answer needs to preserve. If a question asks for one fact, then a single value can settle it: for example, the number listed in a directory. If a question asks about responses across individuals, then the answer needs to be based on data that could take different values from one individual to another.
To see the difference, imagine two possible sets of study-day responses for six learners. In the first set, every learner reports 2 days. In the second, the responses are 0, 1, 2, 3, 4, and 2 days. These are different patterns even though both contain six responses. A question about how many days learners studied must allow the answer to reflect the pattern in the responses; one learner’s value cannot determine the others’ values.
This is a conceptual distinction, not a calculation rule. No statistical formula is needed to decide whether a question anticipates variation. Instead, identify the units whose responses matter and ask whether different responses are possible and relevant to answering the question. If so, the question is statistical. If the answer is one fact regardless of variation among individuals, it is fixed-answer.
What Does Not Determine the Classification
| Feature | Does it make a question statistical? | Why? |
|---|---|---|
| The answer is not yet known | No | An unknown answer can still be one fixed fact. |
| Records must be checked | No | Looking in several places may establish one answer without examining variation across individuals. |
| The answer is numerical | No | A number can be a fixed total or a response that varies across individuals. |
| Responses actually differ | Not by itself | The question is statistical because it anticipates relevant variation, not because a particular sample happened to vary. |
A useful check is to imagine that the data have not yet been collected. Could the individuals give different responses, and would the question need to account for those differences? If yes, the question is statistical. This check avoids classifying questions based only on how much work is needed, whether the response is numerical, or whether the observed answers eventually turn out to match.
Common Mistakes
- Equating uncertainty with statistical thinking. “I do not know how many rooms are listed” does not make the directory question statistical. The answer can still be one fixed count.
- Equating a number with a statistical answer. A total such as 14 rooms is numerical but fixed-answer. A set of wait times is also numerical, but it can vary across visitors.
- Classifying after seeing the data. If everyone in a group gives the same response, the question may still be statistical. Classification depends on what the question anticipates, not just on what happened in one set of observations.
- Treating any data collection as statistical. A person might inspect multiple documents to establish one factual answer. The number of documents consulted is not the issue; the question must call for accounting for variation across individuals.
- Using one person’s response to answer a question about many people. A single response cannot, by itself, show how responses may differ across a group.
Check Your Understanding
For each question, decide whether it is statistical or fixed-answer, and briefly explain why.
- “How many study spaces are listed in the campus directory?”
- “How many hours did each of several students spend studying?”
- “Does the health form include a question about headache?”
- “Which symptoms do patients report?”
- Suppose all visitors in a hypothetical set report the same wait time. Could the question about visitors’ wait times still be statistical? Explain.