Why this chapter matters
Inference is only as credible as the data process. Good sampling and clear variable definitions prevent misleading conclusions.
What you will learn
- Distinguish population parameters from sample statistics.
- Classify variables as categorical or quantitative and identify their measurement context.
- Recognize common sources of bias and describe better sampling plans.
Understand the core ideas
Inference quality starts with study design and sampling quality, not with later formulas. A population is the full group you want to describe, while a sample is the subset you can actually observe. If the sample is selected with a process that excludes certain people or situations, your estimate can miss the population value in a systematic direction. This is bias, and it cannot be fixed by adding more observations from the same flawed process. Variable definitions matter too. If one observer labels borderline cases differently from another observer, measurement inconsistency enters the data and inflates noise or creates fake differences.
A practical workflow is to define target population boundaries, define inclusion rules, and choose a sampling method before collection begins. Random and stratified approaches are often preferred because they reduce predictable overrepresentation of easy-to-reach respondents. Study design then sets claim limits. Observational studies can estimate patterns and associations, but causal statements need stronger design elements, such as controlled assignment. Reliable statistics therefore begin with transparent sampling logic and operational definitions that can be repeated consistently by different people.
Key terms
- Population
- The entire group about which a question is asked or a conclusion is intended.
- Sample
- The observed subset of the population used to produce estimates.
- Parameter
- A numerical characteristic of the population, usually unknown and fixed.
- Bias
- A systematic error from design or measurement that pushes estimates away from truth.
Campus study-hours estimate
A college has 2,000 students. You randomly sample 80 students to estimate average weekly study hours, and the sample mean is 14.2.
- State the population as all 2,000 students and the sample as the 80 selected students.
- Identify 14.2 as a sample statistic, not a population parameter.
- Note that another random sample of 80 would produce a different sample mean.
- Conclude that uncertainty remains even with random sampling because only part of the population was observed.
A common misconception
Claim: A larger sample always fixes a bad sample design.
Correction: A larger sample can reduce random noise, but it does not remove systematic bias from nonrandom selection or inconsistent measurement.
Lessons in this chapter
- Population, sample, and variablesSet up a statistical study using precise language for who and what is measured.
- Sampling methods and biasCompare random, stratified, cluster, and convenience samples and their risks. Read the full guide →
- Observational studies versus experimentsIdentify when a design supports association only or supports causal claims.
- Data quality checksScreen for missing values, outliers, and wording effects before analysis.
Study task
Chapter checkpoint
Why does a convenience sample weaken inference about a full population?
Because participants are chosen by ease of access, not random selection, so the sample can systematically differ from the population.