Why “Who?” and “When?” Matter
A statistical question can invite different answers if it leaves unclear which people it concerns or what period the answer should describe. “How often are students absent?” sounds understandable, but it does not specify which students, or whether “often” refers to a week, a term, or a full school year. Two people could answer it using different groups and periods, then disagree even if both calculated correctly.
A well-scoped question makes those boundaries visible. “Who?” identifies the individuals being described, such as students enrolled at one school or patients discharged from a clinic. “When?” identifies the period relevant to the information, such as a named month or the first seven days after an event. Getting these boundaries clear does not, by itself, decide how a concept such as attendance or a symptom will be recorded. It does make clear whose records and which dates a later measurement must cover.
Make the Group of Individuals Explicit
Naming a broad category of people is often not enough. “Students” might mean every student in a district, students at one school, or students in a particular grade. “Patients” might mean everyone who visited a clinic, or only patients who experienced a specified event there. A useful question identifies the group with enough detail to distinguish the intended individuals from nearby groups that are not included.
This does not require listing every individual by name. It means using a description with a clear boundary: for instance, students enrolled at a named school on a stated date, or adult patients discharged from a named unit during a stated month. If membership can change during the period, the question should make clear which rule determines inclusion. A rule based on status at the start of the period differs from one based on being present at any time during it.
The observational unit also matters. A question about how many students were absent concerns students, even if the records are organized by school day. A question about the number of student absences on each day would instead describe days. The same underlying records can be arranged to answer different questions, but the unit being counted or summarized must match the wording.
Worked Example: Scoping a Student Attendance Question
Suppose a fictional high school wants to ask, “How many days were students absent in October?” The school has records for 120 students who were enrolled there on October 1. The attendance question needs a clear group and period before a summary can be interpreted.
Who? Specify the 120 students enrolled at the school on October 1. This fixes membership for the question rather than leaving “students” open to mean all students in the district or only those present on a particular day.
When? Specify the school days from October 1 through October 31 of the stated year. Naming the year and the date limits prevents “October” from being confused with another year or with a different school reporting period.
Unit and summary. The individuals are students; the requested quantity can be the number of their recorded absent days during that period. Suppose, purely hypothetically, that the records show 186 absent student-days across the 120 students. The average number of absent days per student is
The result describes an average for the stated group and October period. It does not mean that each student missed 1.55 days; individual students have their own records, and the average summarizes them. If the question instead included students who enrolled at any point in October, it could involve a different group and a different denominator. The original numerical calculation might then fail to answer the revised question.
Make the Time Period Precise
A time reference can be a fixed calendar interval, such as October 1 through October 31, or a duration measured from an event, such as the first seven days after a patient is discharged. These are different forms of “when.” A fixed interval gives everyone the same calendar boundaries. An event-based interval gives each person a period anchored to that person’s own event.
Phrases such as “recently,” “during the school year,” and “after treatment” may be too vague for a statistical question unless the intended dates or duration are understood. “After treatment,” for example, could mean the first day, the next week, or a much longer follow-up. “Last month” can also be ambiguous if the date on which it is asked is not recorded. Naming dates, a duration, or a clear event boundary makes the intended period inspectable.
A time period can also affect what counts as an opportunity for an observation. In an attendance question, October contains school days and non-school days; a question about absences on school days should not quietly treat weekends as days when students could have attended school. In a symptom question, a patient who is followed for seven days has a different observation window from one whose information covers only the first day. The question should state the intended window rather than leave readers to supply one.
Worked Example: Fixing the Window for Patient Symptoms
Imagine a fictional clinic asking, “How many patients had nausea after discharge?” The clinic’s records include 60 adult patients discharged from one unit during May. To make the question interpretable, specify both the individuals and the symptom period.
Who? The question concerns the 60 adult patients discharged from the named unit during May. It does not refer to every patient who visited the clinic during May, because a visit without discharge does not satisfy the stated group description.
When? Suppose the intended period is the seven days beginning on the calendar day after each patient’s discharge. This is an event-based window: the dates differ from patient to patient, but the duration and its start rule are the same.
Assume, hypothetically, the records indicate that 18 of the 60 patients had nausea recorded during their specified seven-day window. The proportion for this group and period is
This result means that 30.0% of the 60 specified patients had nausea recorded in the seven-day period after discharge, under the assumed records. It does not describe all clinic patients, and it does not describe symptoms over an unlimited period after discharge. The question’s scope states who and when; a separate measurement plan would still need to specify how nausea is identified and recorded.
Why the Boundaries Change the Answer
The importance of “who” and “when” can be seen directly in a calculation. Suppose one summary is an average across students. Its denominator is the number of students in the stated group. Changing the group can change both that denominator and the total being summarized. Changing the time interval can also change the total, because more or fewer dates contribute observations.
For example, if 186 absent student-days are recorded for 120 students during October, the average is \(186/120=1.55\) absent days per student. If the intended group instead contains 150 students, a valid average for that group would require the absent-day total for those 150 students; dividing 186 by 150 would be justified only if 186 were their total. Similarly, a total for October cannot automatically answer a question about the full term. The arithmetic is meaningful only when its numerator and denominator correspond to the stated group and period.
This gives a simple check for a question and its eventual answer: identify the individuals that contribute to the summary, identify the dates or event-relative window that contribute to the records, and verify that the calculation uses exactly those boundaries. This check does not determine whether the data collection or measurement is otherwise suitable. It prevents a more basic mismatch in which the answer refers to a different set of people or dates than the question.
Worked Example: Distinguishing Two Symptom Periods
Consider a fictional group of 40 adult patients discharged from a unit on June 10. A question asks how many had a particular symptom “soon after discharge.” That phrase has no fixed time boundary, so the same group could produce different answers depending on the intended period.
Suppose the intended period is the first 24 hours after discharge, and 12 of the 40 patients have the symptom recorded in that window. The proportion is
If the question instead concerns the first seven days after discharge, suppose 21 of the same 40 patients have the symptom recorded during that longer window. The proportion for that period is
Both calculations use the same 40 patients, but the time boundaries differ, so the results answer different questions. The seven-day period includes more time in which the symptom could be recorded; it is not a correction to the 24-hour answer. The results also do not establish how the symptom was assessed. For the scope question, the essential point is that the period must be stated before either percentage can be interpreted.
A Practical Scoping Check
Before treating a statistical question as ready to answer, read it for boundaries rather than only for its general topic. A compact review can be done in three steps:
Replace broad labels such as “students” or “patients” with a group description that identifies who is included.
State the calendar dates or the duration and event that define the observation period.
Check that the question and any count, average, or proportion describe the same individuals over the same period.
This review is especially useful when a question will be answered from records collected for another purpose. A database may contain many dates, people, and types of records, but its contents do not decide which of them the question intends to describe. The scope should be stated in the question, not inferred after seeing a convenient subset of the data.
Common Mistakes
- Leaving the group broad. “How many students were absent?” could refer to a grade, school, or district. Specify which students the answer should describe.
- Using an undefined time phrase. “After treatment” or “recently” may cover different periods for different readers. State dates or an event-relative duration.
- Changing the observational unit without noticing. A record for each student answers a student-level question; a record for each school day addresses days. Say which units the question concerns.
- Using a denominator that does not match the question. A proportion among discharged patients is not a proportion among all clinic visitors. Include only the individuals defined by the question, and use the appropriate count.
- Assuming a clear period defines the measurement. “Nausea during the seven days after discharge” specifies when, but not how nausea is assessed. Keep the scope clear and treat the measurement rule as a separate question.
- Assuming more than the answer supports. A percentage for one named group and period describes that group and period. It should not be stated as a result for a broader group or a longer time span.
Check Your Understanding
For each question, identify what needs to be specified or corrected before an answer can be interpreted.
- “How many days are students absent?” Name two details about who and when that would make this question more precise.
- A clinic asks about symptoms “after discharge.” Give one clear event-based period that could replace this phrase.
- Why might an average based on students enrolled on October 1 fail to answer a question about everyone who attended the school at any point in October?
- A report gives the percentage of patients with a recorded symptom in the first 24 hours after discharge. What time period does this result describe, and what does it not automatically describe?
- A school has records for each student and for each school day. Give an example of a question whose observational unit is a student, and one whose observational unit is a day.