Who or What Does Each Row Represent?
Before interpreting a data table, ask two related questions: who or what is the question about, and what does one row represent? The answers often match. If a campus survey has one completed questionnaire per student, then each row represents a student, and the students are the individuals the survey concerns. But they need not match. A table can have one row per household while describing the people in those households, or one row per clinic visit while describing patients who may have made several visits.
This distinction matters because a row count is a count of records, not automatically a count of people or other entities of interest. Treating those counts as interchangeable can lead to incorrect descriptions of a study, misleading summaries, or conclusions that do not answer the intended question.
Statistical writing does not use the terms individual, case, and observational unit identically in every context. In many introductory examples, they refer to the same entity because each row describes one person or object. It is useful to notice the row-level meaning of case and ask explicitly what that row represents, rather than assuming that every row is a different person.
When the Row and the Individual Match
Suppose a university sends a campus dining survey to 120 randomly selected students. Each student completes one questionnaire about their own dining experiences, and the data table contains 120 rows, one per returned questionnaire. If each row is tied to one distinct respondent, then the cases and observational units are the students. The individuals the survey concerns are also those students, so the number of rows and the number of represented students are both 120.
This is a convenient arrangement, but it depends on how information was collected and organized. For example, if a student completes the survey twice, there may be two rows for that student. If one representative answers on behalf of a residence hall, a row may describe a group rather than one student. The table’s row structure should be checked rather than inferred from the subject of the survey.
Worked Example: A Campus Dining Survey
In a hypothetical survey, five students are each asked once about their own use of the campus dining hall. The completed table has five rows, one for each distinct respondent: Maya, Luis, Chen, Aisha, and Noah. Identify the individuals, cases, and observational units.
Individuals. The question concerns the five students who responded: Maya, Luis, Chen, Aisha, and Noah.
Cases. There are five cases because the table has five respondent records.
Observational units. Each case represents one student, so the observational units are the five students.
Here the row count equals the number of students because each student contributes exactly one row and each row represents exactly one distinct student. That conclusion relies on those facts; it does not follow merely from the phrase “student survey.”
When a Row Represents a Group
A table may contain one row per group even when the question is ultimately about the group’s members. For example, a health department could record the number of people in each household who use a local clinic. The cases would be households, while the people whose clinic use is of interest would be household members. A household-level row summarizes information about its members; it is not an individual row for each member.
The difference between these levels becomes especially important when counting. If every household has at least one member, the number of household rows is no greater than the number of people represented. The two counts are equal only when every household contains exactly one person; the household-row count is smaller if at least one household contains multiple people.
Worked Example: Household Rows and Clinic Use
Imagine a hypothetical health survey with one row for each of four households. The households contain 2, 3, 1, and 4 people, respectively. The table reports that 1, 2, 0, and 3 people in those households use a particular clinic. Identify the cases, the people described, and the number of clinic users.
Cases and observational units. There are four cases, and each represents one household. The observational unit at the row level is therefore the household.
People described. The total number of household members is \(2+3+1+4=10\). If the question concerns clinic use among household members, the individuals of interest are all 10 people, of whom 6 use the clinic. The total number of users is \(1+2+0+3=6\).
If instead the question concerns only the people who use the clinic, then there are 6 individuals of interest. Those six people are not the same count as all household members; the other four household members do not use the clinic.
Why the row count is smaller here. Let \(m_j\) be the number of members in household \(j\), for \(j=1,2,3,4\). Each household is nonempty, so each \(m_j\) is at least 1. Thus the total number of people is \(m_1+m_2+m_3+m_4\), which is at least 4. Since the household sizes are 2, 3, 1, and 4, the total is 10, strictly greater than the four household rows. In general, the totals are equal only if every household has one member; any household with more than one member makes the people total larger than the row count.
The four rows are still useful, but they do not give ten separate person-level cases. A table organized by household and a question about people are at different levels of observation.
When One Individual Has Several Rows
The mismatch can also run in the other direction: a data set may have several rows for one individual. Health records often use one row per visit. If a patient attends a clinic three times, that patient may appear in three separate visit records. In that table, the observational unit is the visit, not necessarily the patient.
This structure is appropriate for questions about visits, such as how many visits occurred during a period. It does not mean that every row is a different patient. For a question about patients, the number of distinct patients must be identified separately from the number of visit records. Repeated rows can also affect summaries: patients with more visits contribute more records than patients with fewer visits.
Worked Example: Clinic Visits Versus Patients
In a hypothetical clinic log, five rows represent five visits. Patient A appears twice, Patient B appears once, and Patient C appears twice. How many visit cases and how many distinct patients are represented?
Visit cases. There are five cases because there are five visit records. The observational unit for this table is a clinic visit.
Distinct patients. The patients represented are A, B, and C, so there are three distinct patients. The five rows do not describe five different people because A and C each have two visits.
If the question is “How many visits are recorded?”, the answer is five. If the question is “How many patients are represented?”, the answer is three. The correct count depends on the unit named in the question.
A Reliable Way to Identify the Units
When a study description is brief, work from the question and the data collection process rather than guessing from the table’s subject. The following sequence separates the concepts.
Name the people or other entities the question is intended to describe. Be specific: for example, “all household members in the surveyed households” rather than just “the survey.”
Use the study’s recording procedure. A row might represent one student, one household, one school, or one clinic visit.
Ask whether each row corresponds to one distinct individual of interest, summarizes multiple individuals, or records an additional observation for an individual already represented.
Count rows for cases, distinct entities for individuals, and the relevant represented units for observational units. Do not substitute one count for another without checking the relationship.
The key question is not simply “How many rows are there?” It is “How many rows are there, and what does one row stand for?” Once that is clear, the appropriate count follows from the study’s design.
Why the Distinction Matters
The unit of observation determines what a data row can tell us. A household-level table can describe household totals, but it does not automatically reveal the individual circumstances of every member. A visit-level table can describe recorded visits, but it does not automatically give each patient equal representation if some patients have more visits than others.
This matters when a conclusion is meant to describe a particular group. If the question is about people, an analysis of household summaries may not answer it in the same way as person-level records. If the question is about patients, counting visits as if they were distinct patients overstates how many people are represented. Stating both the target and the row-level unit makes these limitations visible.
Common Mistakes
- Assuming every row is a person. A household row may summarize several people, and a visit row may be one of several records for the same patient. Check the row’s meaning before counting people.
- Using a group total as a person count. Four household rows do not imply four people. In the example, the four households contained 10 people, of whom 6 used the clinic.
- Changing the target without changing the count. “All household members” refers to 10 people in the example; “household members who use the clinic” refers to 6. Those are different sets of individuals.
- Calling repeated records distinct individuals. Five visit records can represent only three patients. Count distinct patients when the question asks about patients.
- Using terms without naming the unit. Instead of writing “there are five observations,” state whether those are five students, five households, or five visits.
Check Your Understanding
For each question, identify the target individuals or entities and explain what a row represents.
- A campus survey has one response from each of 80 distinct students, with one row per response. What are the cases and observational units? How many students are represented?
- A health table has one row per household and lists the number of household members who use a clinic. If the households contain 2, 1, and 4 people, how many people are described, and how many household cases are there?
- A clinic log contains 12 visit records from 7 distinct patients. How many cases are in the log, and how many patients are represented?
- A survey has one row per residence hall, with each row summarizing responses from students living there. Is the row-level observational unit a student or a residence hall? Explain.
- Why is the statement “the data set has 20 rows, so it describes 20 people” not always justified?