Introducing Statistics: Population, Sample, Parameter, and Statistic
Introducing statistics begins with the structure of a statistical question: who or what is being studied, what variables are recorded, what population is targeted, what sample supplies the data, and which sample statistic is used to learn about an unknown population parameter.
Introducing statistics through the data-to-conclusion chain
A useful first model is: investigative question → population or process → observational units → variables → data-collection design → sample data → statistic → interpretation → possible inference about a parameter. Every later AP Statistics method sits somewhere on this chain. If the early labels are wrong, a technically correct calculation can answer the wrong question.
| Term | Meaning | Fast diagnostic question |
|---|---|---|
| Population | The full group or process the question is about | Who or what should the final conclusion describe? |
| Sample | The units actually observed from the population | Which units supplied the recorded data? |
| Parameter | A numerical characteristic of the population | What fixed but usually unknown quantity is the study targeting? |
| Statistic | A numerical summary calculated from the sample | What quantity can be computed from the observed sample? |
| Observational unit | One entity on which variables are recorded | What does one row of the dataset represent? |
| Variable | A characteristic recorded for each unit | What can differ from one observational unit to another? |
Fifteen foundational decisions that prevent later statistics errors
Population versus sample
Population versus sample. A population is the full collection about which the statistical question is posed; a sample is the subset actually observed The evidence to inspect is the wording that defines who or what the conclusion should describe and the list of units actually measured. students sometimes call the largest number in a problem the population even when it describes something else The practical decision is write the population in words before writing the sample size. The population can be people, transactions, manufactured items, days, locations, or repeated process outcomes.
For a written audit of population versus sample, make the evidence visible before deciding whether the material is ready to use. Record the wording that defines who or what the conclusion should describe and the list of units actually measured. Then write one sentence explaining why the decision follows from those details, and one sentence naming what would change the decision. The population can be people, transactions, manufactured items, days, locations, or repeated process outcomes. This extra step matters because students sometimes call the largest number in a problem the population even when it describes something else The final action should therefore be specific: write the population in words before writing the sample size
- Evidence: the wording that defines who or what the conclusion should describe and the list of units actually measured
- Risk: students sometimes call the largest number in a problem the population even when it describes something else
- Decision: write the population in words before writing the sample size
Parameter versus statistic
a parameter summarizes a population while a statistic is computed from sample data For Parameter versus statistic, do not decide from the label alone; examine whether the quantity is fixed but unknown for the population or calculated from observed sample values. name the population target first and then identify the sample statistic used to estimate it symbols such as p and p-hat can be swapped because both are proportions The distinction is the foundation of inference because sampling distributions describe how statistics behave around parameters.
A useful checkpoint for parameter versus statistic is whether another student or teacher could reproduce the decision from the evidence recorded. The record should include whether the quantity is fixed but unknown for the population or calculated from observed sample values. If the page, book, activity, or exam note cannot supply that information, treat the gap as unresolved rather than filling it with an assumption. symbols such as p and p-hat can be swapped because both are proportions In that situation, name the population target first and then identify the sample statistic used to estimate it The distinction is the foundation of inference because sampling distributions describe how statistics behave around parameters.
- Evidence: whether the quantity is fixed but unknown for the population or calculated from observed sample values
- Risk: symbols such as p and p-hat can be swapped because both are proportions
- Decision: name the population target first and then identify the sample statistic used to estimate it
Observational unit
Observational unit becomes useful when it changes a concrete study decision. the observational unit is the entity on which variables are recorded Start with one row of the dataset and the real-world object that row represents; then ask what one row means before classifying any variable. A common failure mode is this: students may name the variable itself as the observational unit A row could represent a person, school, county, transaction, tree, machine cycle, or experimental plot.
Turn observational unit into a two-column note. In the first column, write the observable facts: one row of the dataset and the real-world object that row represents. In the second, write the implication for study or instruction. The implication should be ask what one row means before classifying any variable. This format exposes weak reasoning quickly, because students may name the variable itself as the observational unit A row could represent a person, school, county, transaction, tree, machine cycle, or experimental plot. It also creates a reusable record for the next review cycle without copying generic advice.
- Evidence: one row of the dataset and the real-world object that row represents
- Risk: students may name the variable itself as the observational unit
- Decision: ask what one row means before classifying any variable
Variable
a variable is a characteristic that can differ from one observational unit to another The checkpoint for Variable is the recorded field, its possible values, and what those values mean. If that checkpoint is satisfied, describe the variable in words and identify whether it is categorical or quantitative. If it is ignored, dataset column names can be mistaken for units or populations Clear variable definitions prevent later errors in graph choice and interpretation.
When reviewing variable, ask what evidence would persuade you that the current plan is correct and what evidence would make you revise it. Start from the recorded field, its possible values, and what those values mean. The most defensible next move is to describe the variable in words and identify whether it is categorical or quantitative. Do not let familiarity substitute for verification, because dataset column names can be mistaken for units or populations Clear variable definitions prevent later errors in graph choice and interpretation. The purpose of the check is to improve a real decision, not to accumulate another note.
- Evidence: the recorded field, its possible values, and what those values mean
- Risk: dataset column names can be mistaken for units or populations
- Decision: describe the variable in words and identify whether it is categorical or quantitative
Categorical variable
Use Categorical variable as an audit question rather than a heading to memorize. categorical variables place units into groups or labels rather than measuring a numeric amount Look for the meaning of the values, even when categories are stored with numeric codes. From there, classify by meaning, not storage format. The warning sign is that numbers such as 1,2,3 can be treated as quantitative merely because arithmetic is possible in software A mean of category codes is usually meaningless unless the codes genuinely represent a quantitative scale.
For a written audit of categorical variable, make the evidence visible before deciding whether the material is ready to use. Record the meaning of the values, even when categories are stored with numeric codes. Then write one sentence explaining why the decision follows from those details, and one sentence naming what would change the decision. A mean of category codes is usually meaningless unless the codes genuinely represent a quantitative scale. This extra step matters because numbers such as 1,2,3 can be treated as quantitative merely because arithmetic is possible in software The final action should therefore be specific: classify by meaning, not storage format
- Evidence: the meaning of the values, even when categories are stored with numeric codes
- Risk: numbers such as 1,2,3 can be treated as quantitative merely because arithmetic is possible in software
- Decision: classify by meaning, not storage format
Quantitative variable
quantitative variables record numerical amounts for which arithmetic differences have meaning In the case of Quantitative variable, the strongest evidence is measurement units, scale, possible zero point, and whether comparisons such as difference or ratio are sensible. That evidence should lead to this action: ask whether adding or averaging the values answers a meaningful question. By contrast, every value containing digits can be labeled quantitative even when it is an ID number Examples include time, distance, mass, income, temperature, and count outcomes.
A useful checkpoint for quantitative variable is whether another student or teacher could reproduce the decision from the evidence recorded. The record should include measurement units, scale, possible zero point, and whether comparisons such as difference or ratio are sensible. If the page, book, activity, or exam note cannot supply that information, treat the gap as unresolved rather than filling it with an assumption. every value containing digits can be labeled quantitative even when it is an ID number In that situation, ask whether adding or averaging the values answers a meaningful question Examples include time, distance, mass, income, temperature, and count outcomes.
- Evidence: measurement units, scale, possible zero point, and whether comparisons such as difference or ratio are sensible
- Risk: every value containing digits can be labeled quantitative even when it is an ID number
- Decision: ask whether adding or averaging the values answers a meaningful question
Investigative question
Investigative question should be checked in context. a statistical investigative question anticipates variability and can be answered with data The decisive details are population or process, variables, comparison or relationship, and the kind of evidence needed. a question such as “Is tutoring good?” is too vague to determine data or analysis A sound response is to rewrite vague questions until the target and variables are explicit. Good questions make later choices about sampling, graphing, and inference much easier.
Turn investigative question into a two-column note. In the first column, write the observable facts: population or process, variables, comparison or relationship, and the kind of evidence needed. In the second, write the implication for study or instruction. The implication should be rewrite vague questions until the target and variables are explicit. This format exposes weak reasoning quickly, because a question such as “Is tutoring good?” is too vague to determine data or analysis Good questions make later choices about sampling, graphing, and inference much easier. It also creates a reusable record for the next review cycle without copying generic advice.
- Evidence: population or process, variables, comparison or relationship, and the kind of evidence needed
- Risk: a question such as “Is tutoring good?” is too vague to determine data or analysis
- Decision: rewrite vague questions until the target and variables are explicit
Descriptive versus inferential goal
descriptive statistics summarize observed data while inference uses sample evidence to learn about a broader population or process Treat Descriptive versus inferential goal as a verification task: identify whether the conclusion is limited to the recorded sample or extends beyond it, and then state the scope of the desired conclusion before selecting methods. This avoids a frequent mistake in which a numerical summary can be presented as proof about a population without a defensible sampling design Inference requires attention to how the data were produced, not just what calculations are available.
When reviewing descriptive versus inferential goal, ask what evidence would persuade you that the current plan is correct and what evidence would make you revise it. Start from whether the conclusion is limited to the recorded sample or extends beyond it. The most defensible next move is to state the scope of the desired conclusion before selecting methods. Do not let familiarity substitute for verification, because a numerical summary can be presented as proof about a population without a defensible sampling design Inference requires attention to how the data were produced, not just what calculations are available. The purpose of the check is to improve a real decision, not to accumulate another note.
- Evidence: whether the conclusion is limited to the recorded sample or extends beyond it
- Risk: a numerical summary can be presented as proof about a population without a defensible sampling design
- Decision: state the scope of the desired conclusion before selecting methods
Census versus sample
Census versus sample. a census attempts to observe every member of a population, while a sample observes only part The evidence to inspect is coverage, cost, timeliness, measurement quality, and whether the population changes during data collection. a census can be assumed automatically better even when nonresponse or measurement error remains The practical decision is compare total-error sources instead of judging only by sample size. Well-designed samples can be more practical and sometimes more accurate than a poorly executed census.
For a written audit of census versus sample, make the evidence visible before deciding whether the material is ready to use. Record coverage, cost, timeliness, measurement quality, and whether the population changes during data collection. Then write one sentence explaining why the decision follows from those details, and one sentence naming what would change the decision. Well-designed samples can be more practical and sometimes more accurate than a poorly executed census. This extra step matters because a census can be assumed automatically better even when nonresponse or measurement error remains The final action should therefore be specific: compare total-error sources instead of judging only by sample size
- Evidence: coverage, cost, timeliness, measurement quality, and whether the population changes during data collection
- Risk: a census can be assumed automatically better even when nonresponse or measurement error remains
- Decision: compare total-error sources instead of judging only by sample size
Sampling variability
different random samples from the same population generally produce different statistics For Sampling variability, do not decide from the label alone; examine the repeated-sampling distribution of the statistic and how its spread changes with sample size. expect ordinary sample-to-sample variation and quantify it through standard error or simulation when appropriate students may think two different sample proportions prove one study is wrong Sampling variability is not a mistake; it is a predictable consequence of using a subset to learn about a population.
A useful checkpoint for sampling variability is whether another student or teacher could reproduce the decision from the evidence recorded. The record should include the repeated-sampling distribution of the statistic and how its spread changes with sample size. If the page, book, activity, or exam note cannot supply that information, treat the gap as unresolved rather than filling it with an assumption. students may think two different sample proportions prove one study is wrong In that situation, expect ordinary sample-to-sample variation and quantify it through standard error or simulation when appropriate Sampling variability is not a mistake; it is a predictable consequence of using a subset to learn about a population.
- Evidence: the repeated-sampling distribution of the statistic and how its spread changes with sample size
- Risk: students may think two different sample proportions prove one study is wrong
- Decision: expect ordinary sample-to-sample variation and quantify it through standard error or simulation when appropriate
Bias versus variability
Bias versus variability becomes useful when it changes a concrete study decision. bias is a systematic tendency of a method while variability is natural sample-to-sample fluctuation Start with the data-collection mechanism and the repeated behavior of the estimator; then separate method bias from random sampling variability when diagnosing an estimate. A common failure mode is this: a large sample can be assumed to fix selection bias Increasing n can reduce standard error but cannot rescue a systematically unrepresentative sampling process.
Turn bias versus variability into a two-column note. In the first column, write the observable facts: the data-collection mechanism and the repeated behavior of the estimator. In the second, write the implication for study or instruction. The implication should be separate method bias from random sampling variability when diagnosing an estimate. This format exposes weak reasoning quickly, because a large sample can be assumed to fix selection bias Increasing n can reduce standard error but cannot rescue a systematically unrepresentative sampling process. It also creates a reusable record for the next review cycle without copying generic advice.
- Evidence: the data-collection mechanism and the repeated behavior of the estimator
- Risk: a large sample can be assumed to fix selection bias
- Decision: separate method bias from random sampling variability when diagnosing an estimate
Random sample
random sampling uses a chance mechanism to select units from a defined frame The checkpoint for Random sample is the population frame, randomization mechanism, and whether each selection probability is appropriate. If that checkpoint is satisfied, describe the chance mechanism explicitly. If it is ignored, “random” can be claimed merely because the researcher chose units without a written plan Random selection is primarily connected to representativeness and generalization, not to causal treatment comparisons.
When reviewing random sample, ask what evidence would persuade you that the current plan is correct and what evidence would make you revise it. Start from the population frame, randomization mechanism, and whether each selection probability is appropriate. The most defensible next move is to describe the chance mechanism explicitly. Do not let familiarity substitute for verification, because “random” can be claimed merely because the researcher chose units without a written plan Random selection is primarily connected to representativeness and generalization, not to causal treatment comparisons. The purpose of the check is to improve a real decision, not to accumulate another note.
- Evidence: the population frame, randomization mechanism, and whether each selection probability is appropriate
- Risk: “random” can be claimed merely because the researcher chose units without a written plan
- Decision: describe the chance mechanism explicitly
Random assignment
Use Random assignment as an audit question rather than a heading to memorize. random assignment uses chance to place experimental units into treatment conditions Look for treatment labels, assignment mechanism, control condition, and response measurement. From there, state whether chance selected units from a population or assigned treatments to recruited units. The warning sign is that random assignment is sometimes confused with random sampling Random assignment supports causal comparison; it does not by itself make volunteers representative of a broader population.
For a written audit of random assignment, make the evidence visible before deciding whether the material is ready to use. Record treatment labels, assignment mechanism, control condition, and response measurement. Then write one sentence explaining why the decision follows from those details, and one sentence naming what would change the decision. Random assignment supports causal comparison; it does not by itself make volunteers representative of a broader population. This extra step matters because random assignment is sometimes confused with random sampling The final action should therefore be specific: state whether chance selected units from a population or assigned treatments to recruited units
- Evidence: treatment labels, assignment mechanism, control condition, and response measurement
- Risk: random assignment is sometimes confused with random sampling
- Decision: state whether chance selected units from a population or assigned treatments to recruited units
Association versus causation
an observed association means variables move together in the data; causation requires stronger design evidence In the case of Association versus causation, the strongest evidence is study design, possible confounding, temporal order, and random assignment when an experiment is feasible. That evidence should lead to this action: match causal language to the data-production design. By contrast, a strong correlation or small p-value can be described as proof of cause Statistical significance does not convert an observational study into a randomized experiment.
A useful checkpoint for association versus causation is whether another student or teacher could reproduce the decision from the evidence recorded. The record should include study design, possible confounding, temporal order, and random assignment when an experiment is feasible. If the page, book, activity, or exam note cannot supply that information, treat the gap as unresolved rather than filling it with an assumption. a strong correlation or small p-value can be described as proof of cause In that situation, match causal language to the data-production design Statistical significance does not convert an observational study into a randomized experiment.
- Evidence: study design, possible confounding, temporal order, and random assignment when an experiment is feasible
- Risk: a strong correlation or small p-value can be described as proof of cause
- Decision: match causal language to the data-production design
Context in conclusions
Context in conclusions should be checked in context. statistics answers are incomplete when the numerical result is detached from the variables and population The decisive details are units, group labels, parameter, direction, and the exact claim being evaluated. students may write “the mean is higher” without saying whose mean or by how much A sound response is to attach every interpretation to the original scenario. Context is not decorative wording; it specifies what the number means and what conclusion is justified.
Turn context in conclusions into a two-column note. In the first column, write the observable facts: units, group labels, parameter, direction, and the exact claim being evaluated. In the second, write the implication for study or instruction. The implication should be attach every interpretation to the original scenario. This format exposes weak reasoning quickly, because students may write “the mean is higher” without saying whose mean or by how much Context is not decorative wording; it specifies what the number means and what conclusion is justified. It also creates a reusable record for the next review cycle without copying generic advice.
- Evidence: units, group labels, parameter, direction, and the exact claim being evaluated
- Risk: students may write “the mean is higher” without saying whose mean or by how much
- Decision: attach every interpretation to the original scenario
16 original multiple-choice checks for introducing statistics
These items focus on the vocabulary and reasoning that support every later unit. Answer them by identifying the statistical object first—population, sample, parameter, statistic, observational unit, variable, or design role—before looking for a formula.
Question 1. Population and sample
A university has 18,000 undergraduates. Researchers randomly select 300 and record weekly study hours. Which is the sample?
Answer: B
The sample is the set of 300 observational units actually selected and measured.
Question 2. Parameter and statistic
In the same university study, the sample mean study time is 11.4 hours. Which quantity is a statistic?
Answer: B
The 11.4-hour mean is computed from the sample, so it is a statistic. The corresponding population mean is a parameter.
Question 3. Observational unit
A dataset has one row for each restaurant inspection and columns for date, score, restaurant type, and inspector. What is the observational unit?
Answer: A
One row represents one inspection, so an inspection is the observational unit.
Question 4. Categorical variable
Which variable is categorical?
Answer: C
Transportation mode records labels such as car, bus, bicycle, or walking rather than a measured numerical amount.
Question 5. Quantitative variable
Which variable is quantitative?
Answer: D
Rainfall is a numerical measurement with meaningful arithmetic and units.
Question 6. Investigative question
Which question is most clearly statistical?
Answer: C
The commute-time question anticipates variability, identifies groups and a quantitative variable, and can be answered with data.
Question 7. Descriptive goal
A class calculates the median test score for the 28 students who took one quiz and makes no claim beyond those 28 students. This is primarily:
Answer: A
The calculation summarizes the observed class without extending a conclusion to a broader population.
Question 8. Inferential goal
A random sample of county voters is used to estimate the proportion of all county voters supporting a bond issue. The target is:
Answer: B
The goal is to learn about the population proportion using a sample statistic.
Question 9. Census
Which statement about a census is correct?
Answer: B
A census attempts complete population coverage, but it can still have nonresponse, measurement, or processing errors.
Question 10. Sampling variability
Two independent simple random samples from the same population produce proportions 0.48 and 0.53. What is the best first explanation?
Answer: B
Sample-to-sample variation is expected even when both samples are selected correctly from the same population.
Question 11. Bias
A website poll asks visitors to click if they strongly support a policy. The biggest concern is:
Answer: A
People choose whether to respond, and those with strong opinions may be overrepresented.
Question 12. Large sample and bias
A convenience sample grows from 100 to 10,000 participants but uses the same biased recruitment method. What is most accurate?
Answer: B
Increasing sample size can reduce random variability but does not fix a systematically unrepresentative selection process.
Question 13. Random selection
Randomly selecting names from a population list primarily helps with:
Answer: B
Random selection is tied to how well a sample represents the target population, subject to the quality of the frame and response process.
Question 14. Random assignment
Randomly assigning recruited participants to two treatments primarily helps with:
Answer: A
Random assignment balances lurking variables in expectation and supports causal interpretation for the experimental units.
Question 15. Association and causation
An observational study finds that people who sleep more report lower stress. Which conclusion is safest?
Answer: B
Association can be described, but confounding and lack of random assignment prevent a causal conclusion.
Question 16. Context
A sample mean difference is 4.2. What is missing from the bare statement “the difference is 4.2”?
Answer: B
A numerical result needs context so readers know what was subtracted, in what units, and what group or population the result describes.
8 original free-response foundations
Use complete sentences and context. The goal is not length; it is precision about what was observed, what is being estimated, and what the data-production method allows you to conclude.
Free-response 1. Population, sample, parameter, statistic
A transit authority has 42,000 weekday riders. It selects a simple random sample of 250 riders and finds that 61% used a mobile ticket. Identify the population, sample, parameter, and statistic.
Model response
Population: all 42,000 weekday riders in the stated frame. Sample: the 250 selected riders. Parameter: p, the true proportion of all weekday riders who use a mobile ticket. Statistic: p-hat=0.61, the sample proportion among the 250 selected riders.
Free-response 2. Observational units and variables
A dataset has one row for each package shipped by a warehouse and columns for destination region, package mass, delivery days, shipping method, and damage status. Identify the observational unit and classify each variable.
Model response
The observational unit is one shipped package. Destination region, shipping method, and damage status are categorical. Package mass and delivery days are quantitative because numerical differences and summaries are meaningful.
Free-response 3. Write an investigative question
A school wants to study whether commute mode is related to tardiness. Write a statistical investigative question and name the needed variables.
Model response
One suitable question is: Among students at the school this semester, how does the proportion arriving late differ across primary commute modes? Needed variables include commute mode (categorical) and tardiness status or number of late arrivals, with the population and time window clearly defined.
Free-response 4. Descriptive versus inferential scope
A teacher summarizes quiz scores for one AP Statistics class and then claims the same mean applies to all AP Statistics students in the state. Evaluate the two steps.
Model response
Computing the class mean is a valid descriptive summary of the observed class. Extending that mean to all AP Statistics students in the state is not justified unless the class can be defended as a representative sample of that state population. The scope of inference comes from how the data were obtained.
Free-response 5. Bias versus variability
Explain why taking a much larger voluntary-response sample may still give a poor estimate of a population opinion.
Model response
A larger sample can reduce random sampling variability, but voluntary response is a systematic selection mechanism: people who choose to respond may differ from those who do not. If that difference is related to the opinion being measured, the estimate can remain biased even with a very large number of responses.
Free-response 6. Random selection versus assignment
A company randomly samples employees from its workforce and then randomly assigns sampled employees to two training programs. Explain the separate purpose of each random step.
Model response
Random selection is used to make the sampled employees more representative of the workforce and can support generalization if the frame and response process are appropriate. Random assignment creates comparable treatment groups and supports a causal comparison of the two training programs for the experimental units.
Free-response 7. Numeric category codes
A survey stores education level as 1=high school, 2=some college, 3=bachelor’s, 4=graduate. Explain why calculating a mean code can be misleading.
Model response
The digits are ordered category labels, not measured quantities with guaranteed equal spacing. The difference between codes 1 and 2 is not a defined amount comparable to the difference between 3 and 4. Frequencies, proportions, or an appropriate ordered-category analysis are more meaningful than treating the codes as ordinary quantitative measurements.
Free-response 8. Contextual interpretation
A sample from two neighborhoods gives a difference in estimated recycling proportions of 0.12, computed as Neighborhood A minus Neighborhood B. Write a contextual interpretation and state one limitation.
Model response
The sample proportion reporting recycling is 0.12, or 12 percentage points, higher in Neighborhood A than in Neighborhood B for the sampled households. Whether that difference can be generalized to all households depends on how the households were selected and whether nonresponse or coverage problems threaten representativeness.
Introducing statistics: a final self-check before moving deeper into Unit 1
You are ready to move on when you can look at a short study description and immediately identify the observational unit, variables, population, sample, parameter or target, sample statistic, and whether the goal is descriptive or inferential. You should also be able to explain the different jobs of random selection and random assignment without using the two phrases interchangeably.
If one of those steps is slow, return to the matching case above and write a new example from a different setting. Introducing statistics is not a vocabulary detour. These distinctions control graph selection, sampling logic, confidence intervals, significance tests, regression interpretation, and the scope of every conclusion later in the course.
Ten worked mini-cases for identifying the statistical structure
Election-poll mini-case
Election-poll mini-case. A city has 84,000 registered voters. A random sample of 600 voters is asked whether they support a transit bond, and 348 say yes. The evidence to inspect is population: all 84,000 registered voters; sample: 600 selected voters; parameter: p, the population support proportion; statistic: p-hat=348/600=0.58. calling 0.58 the parameter would confuse the observed sample result with the unknown citywide target The practical decision is write the four labels before discussing margin of error or confidence intervals. This case shows why inference begins with a parameter-statistic distinction rather than with a calculator menu.
For a written audit of election-poll mini-case, make the evidence visible before deciding whether the material is ready to use. Record population: all 84,000 registered voters; sample: 600 selected voters; parameter: p, the population support proportion; statistic: p-hat=348/600=0.58. Then write one sentence explaining why the decision follows from those details, and one sentence naming what would change the decision. This case shows why inference begins with a parameter-statistic distinction rather than with a calculator menu. This extra step matters because calling 0.58 the parameter would confuse the observed sample result with the unknown citywide target The final action should therefore be specific: write the four labels before discussing margin of error or confidence intervals
- Evidence: population: all 84,000 registered voters; sample: 600 selected voters; parameter: p, the population support proportion; statistic: p-hat=348/600=0.58
- Risk: calling 0.58 the parameter would confuse the observed sample result with the unknown citywide target
- Decision: write the four labels before discussing margin of error or confidence intervals
Manufacturing mini-case
A plant produces thousands of bearings per day. Quality engineers inspect 120 bearings and record diameter in millimeters and whether each bearing passes tolerance. For Manufacturing mini-case, do not decide from the label alone; examine observational unit: one bearing; diameter: quantitative; pass status: categorical; possible parameters include the population mean diameter or population pass proportion. state the investigative question before deciding which variable and parameter matter one dataset can contain several variables and therefore several possible statistical questions The same sample can support different analyses, but each analysis needs a clearly named target.
A useful checkpoint for manufacturing mini-case is whether another student or teacher could reproduce the decision from the evidence recorded. The record should include observational unit: one bearing; diameter: quantitative; pass status: categorical; possible parameters include the population mean diameter or population pass proportion. If the page, book, activity, or exam note cannot supply that information, treat the gap as unresolved rather than filling it with an assumption. one dataset can contain several variables and therefore several possible statistical questions In that situation, state the investigative question before deciding which variable and parameter matter The same sample can support different analyses, but each analysis needs a clearly named target.
- Evidence: observational unit: one bearing; diameter: quantitative; pass status: categorical; possible parameters include the population mean diameter or population pass proportion
- Risk: one dataset can contain several variables and therefore several possible statistical questions
- Decision: state the investigative question before deciding which variable and parameter matter
School-attendance mini-case
School-attendance mini-case becomes useful when it changes a concrete study decision. A district wants to know whether average absence days differ between students who ride a bus and students who walk. Start with population and time period, commute-mode group, quantitative absence variable, and whether the sampled students represent the district; then treat the result as an association unless the design contains a defensible causal mechanism. A common failure mode is this: the comparison can be described causally even though commute mode was not randomly assigned The observational-unit definition should be one student, not one absence day, if each row records a student’s total absences.
Turn school-attendance mini-case into a two-column note. In the first column, write the observable facts: population and time period, commute-mode group, quantitative absence variable, and whether the sampled students represent the district. In the second, write the implication for study or instruction. The implication should be treat the result as an association unless the design contains a defensible causal mechanism. This format exposes weak reasoning quickly, because the comparison can be described causally even though commute mode was not randomly assigned The observational-unit definition should be one student, not one absence day, if each row records a student’s total absences. It also creates a reusable record for the next review cycle without copying generic advice.
- Evidence: population and time period, commute-mode group, quantitative absence variable, and whether the sampled students represent the district
- Risk: the comparison can be described causally even though commute mode was not randomly assigned
- Decision: treat the result as an association unless the design contains a defensible causal mechanism
Website-experiment mini-case
A website randomly assigns consenting visitors to a blue or green checkout button and records whether a purchase is completed. The checkpoint for Website-experiment mini-case is experimental unit: consenting visitor session under the stated design; treatment: button color; response: purchase completion; random assignment supports a causal treatment comparison. If that checkpoint is satisfied, separate causal validity for the experiment from external generalization to a broader population. If it is ignored, the volunteer visitor pool may be generalized to all internet users without random sampling Random assignment and random sampling answer different questions even when both use chance.
When reviewing website-experiment mini-case, ask what evidence would persuade you that the current plan is correct and what evidence would make you revise it. Start from experimental unit: consenting visitor session under the stated design; treatment: button color; response: purchase completion; random assignment supports a causal treatment comparison. The most defensible next move is to separate causal validity for the experiment from external generalization to a broader population. Do not let familiarity substitute for verification, because the volunteer visitor pool may be generalized to all internet users without random sampling Random assignment and random sampling answer different questions even when both use chance. The purpose of the check is to improve a real decision, not to accumulate another note.
- Evidence: experimental unit: consenting visitor session under the stated design; treatment: button color; response: purchase completion; random assignment supports a causal treatment comparison
- Risk: the volunteer visitor pool may be generalized to all internet users without random sampling
- Decision: separate causal validity for the experiment from external generalization to a broader population
Hospital-wait mini-case
Use Hospital-wait mini-case as an audit question rather than a heading to memorize. A hospital records waiting time for every emergency-department arrival during one weekend. Look for the data may form a census of that weekend’s arrivals but not automatically of all future arrivals; wait time is quantitative and arrival is the observational unit. From there, define the population as the weekend arrivals if that is the actual complete group, or use a sampling model for broader process claims. The warning sign is that calling the weekend data a census can lead to unlimited generalization beyond the defined population Population definitions depend on the investigative question, not on the prestige or size of the dataset.
For a written audit of hospital-wait mini-case, make the evidence visible before deciding whether the material is ready to use. Record the data may form a census of that weekend’s arrivals but not automatically of all future arrivals; wait time is quantitative and arrival is the observational unit. Then write one sentence explaining why the decision follows from those details, and one sentence naming what would change the decision. Population definitions depend on the investigative question, not on the prestige or size of the dataset. This extra step matters because calling the weekend data a census can lead to unlimited generalization beyond the defined population The final action should therefore be specific: define the population as the weekend arrivals if that is the actual complete group, or use a sampling model for broader process claims
- Evidence: the data may form a census of that weekend’s arrivals but not automatically of all future arrivals; wait time is quantitative and arrival is the observational unit
- Risk: calling the weekend data a census can lead to unlimited generalization beyond the defined population
- Decision: define the population as the weekend arrivals if that is the actual complete group, or use a sampling model for broader process claims
App-rating mini-case
An app store displays ratings from users who voluntarily chose to submit a review. In the case of App-rating mini-case, the strongest evidence is sample: submitted reviewers; target population might be all users; statistic: average or proportion from reviewers; key design concern: voluntary response. That evidence should lead to this action: identify who had a chance and motivation to respond before interpreting the sample statistic as a population estimate. By contrast, a huge number of reviews can be treated as automatically representative of all users Large n reduces some random variability but does not erase self-selection.
A useful checkpoint for app-rating mini-case is whether another student or teacher could reproduce the decision from the evidence recorded. The record should include sample: submitted reviewers; target population might be all users; statistic: average or proportion from reviewers; key design concern: voluntary response. If the page, book, activity, or exam note cannot supply that information, treat the gap as unresolved rather than filling it with an assumption. a huge number of reviews can be treated as automatically representative of all users In that situation, identify who had a chance and motivation to respond before interpreting the sample statistic as a population estimate Large n reduces some random variability but does not erase self-selection.
- Evidence: sample: submitted reviewers; target population might be all users; statistic: average or proportion from reviewers; key design concern: voluntary response
- Risk: a huge number of reviews can be treated as automatically representative of all users
- Decision: identify who had a chance and motivation to respond before interpreting the sample statistic as a population estimate
Weather-station mini-case
Weather-station mini-case should be checked in context. A station records daily high temperature for 365 days and a researcher studies the distribution of those values. The decisive details are observational unit: a day; variable: daily high temperature; population depends on whether the target is that year or a broader climate process. 365 observations can be called a random sample of all future days without considering time dependence or the defined target A sound response is to state the time population or process explicitly before extending conclusions. The word “population” can refer to a finite list or a conceptual process, but the target must be clear.
Turn weather-station mini-case into a two-column note. In the first column, write the observable facts: observational unit: a day; variable: daily high temperature; population depends on whether the target is that year or a broader climate process. In the second, write the implication for study or instruction. The implication should be state the time population or process explicitly before extending conclusions. This format exposes weak reasoning quickly, because 365 observations can be called a random sample of all future days without considering time dependence or the defined target The word “population” can refer to a finite list or a conceptual process, but the target must be clear. It also creates a reusable record for the next review cycle without copying generic advice.
- Evidence: observational unit: a day; variable: daily high temperature; population depends on whether the target is that year or a broader climate process
- Risk: 365 observations can be called a random sample of all future days without considering time dependence or the defined target
- Decision: state the time population or process explicitly before extending conclusions
Delivery-route mini-case
A logistics company samples 80 deliveries from each of four regions and records delivery time and whether the package was late. Treat Delivery-route mini-case as a verification task: identify strata: regions; delivery is the observational unit; delivery time is quantitative; late status is categorical; stratification guarantees sampled observations from each region, and then respect the stratified design when estimating a companywide quantity, especially if sampling fractions differ by region. This avoids a frequent mistake in which all 320 deliveries can be analyzed as if they came from a simple random sample with no design structure Sampling design is part of the data, not merely a sentence in the methods section.
When reviewing delivery-route mini-case, ask what evidence would persuade you that the current plan is correct and what evidence would make you revise it. Start from strata: regions; delivery is the observational unit; delivery time is quantitative; late status is categorical; stratification guarantees sampled observations from each region. The most defensible next move is to respect the stratified design when estimating a companywide quantity, especially if sampling fractions differ by region. Do not let familiarity substitute for verification, because all 320 deliveries can be analyzed as if they came from a simple random sample with no design structure Sampling design is part of the data, not merely a sentence in the methods section. The purpose of the check is to improve a real decision, not to accumulate another note.
- Evidence: strata: regions; delivery is the observational unit; delivery time is quantitative; late status is categorical; stratification guarantees sampled observations from each region
- Risk: all 320 deliveries can be analyzed as if they came from a simple random sample with no design structure
- Decision: respect the stratified design when estimating a companywide quantity, especially if sampling fractions differ by region
Paired-fitness mini-case
Paired-fitness mini-case. Twenty participants have resting heart rate measured before and after a six-week program. The evidence to inspect is observational unit for the experiment is one participant; two measurements form a pair; a useful derived variable is after-minus-before heart-rate change. the two columns can be treated as independent samples because they contain separate numbers The practical decision is preserve the pairing and analyze within-person differences when the question targets mean change. Recognizing the observational unit early prevents a later inference-method error.
For a written audit of paired-fitness mini-case, make the evidence visible before deciding whether the material is ready to use. Record observational unit for the experiment is one participant; two measurements form a pair; a useful derived variable is after-minus-before heart-rate change. Then write one sentence explaining why the decision follows from those details, and one sentence naming what would change the decision. Recognizing the observational unit early prevents a later inference-method error. This extra step matters because the two columns can be treated as independent samples because they contain separate numbers The final action should therefore be specific: preserve the pairing and analyze within-person differences when the question targets mean change
- Evidence: observational unit for the experiment is one participant; two measurements form a pair; a useful derived variable is after-minus-before heart-rate change
- Risk: the two columns can be treated as independent samples because they contain separate numbers
- Decision: preserve the pairing and analyze within-person differences when the question targets mean change
Library-survey mini-case
A library surveys every 20th visitor after a random starting point and asks satisfaction on a five-level ordered scale. For Library-survey mini-case, do not decide from the label alone; examine systematic sampling mechanism, visitor as observational unit, satisfaction as an ordered categorical variable, and the target visitor population. decide whether the scale is being treated categorically or quantitatively and justify that choice before summarizing the 1–5 categories can be averaged automatically as if equal numerical spacing were guaranteed Storage as digits does not determine a variable’s statistical type.
A useful checkpoint for library-survey mini-case is whether another student or teacher could reproduce the decision from the evidence recorded. The record should include systematic sampling mechanism, visitor as observational unit, satisfaction as an ordered categorical variable, and the target visitor population. If the page, book, activity, or exam note cannot supply that information, treat the gap as unresolved rather than filling it with an assumption. the 1–5 categories can be averaged automatically as if equal numerical spacing were guaranteed In that situation, decide whether the scale is being treated categorically or quantitatively and justify that choice before summarizing Storage as digits does not determine a variable’s statistical type.
- Evidence: systematic sampling mechanism, visitor as observational unit, satisfaction as an ordered categorical variable, and the target visitor population
- Risk: the 1–5 categories can be averaged automatically as if equal numerical spacing were guaranteed
- Decision: decide whether the scale is being treated categorically or quantitatively and justify that choice before summarizing
