Distinguish populations and samples and evaluate selection methods

Learning goal: Identify a population and a sample, explain how selection can bias a conclusion and distinguish a representative sample from one that merely has a convenient size. Recognize variation among possible random samples.

A population is the complete group about which a statistical question asks. A sample is a subset examined to learn about that population. If the question concerns books on two named shelves on a particular day, those books form the population. If the question concerns every book in a school library, examining those same shelves would instead produce only a sample, with a possible selection problem. Population is defined by the question, not by a universal label attached to an object.

A census examines the entire population as defined. It can avoid sampling variation, but measurements and records can still contain errors. A sample can be practical when examining everyone or everything would take too much time, but conclusions about the wider group depend on how well it represents the population in relevant characteristics.

Use a small model whose whole population is visible: twenty numbered cards, five orange and fifteen blue. Cards 1 through 5 are orange, and cards 6 through 20 are blue. The actual orange share is 5/20 = 1/4, or 25%. The diagram keeps all twenty cards visible so that a sample's result can be compared with the known model population.

A convenience sample of the first four cards contains four orange and no blue, suggesting 100% orange if generalized carelessly. The problem is not a difficult calculation; it is a selection rule tied to where the orange cards are clustered. Repeating that same convenience selection does not remove the bias. A larger sample from only the front of a similarly arranged much larger deck could remain misleading.

One random method numbers every card, places matching number slips in a container, mixes thoroughly and draws the required number without replacement. Each card has the same chance of being selected under the same rule. For a simple random sample of four, each possible four-card subset should have the same chance under the model. Do not select only the numbers that are easy to reach or replace an inconvenient result with a preferred one.

A possible four-card sample is 2, 8, 13 and 19. It contains one orange and three blue, matching the population's 25% orange share. Another possible sample is 3, 5, 10 and 17, containing two orange and two blue, or 50% orange. Both could occur under the random method. A random procedure reduces systematic selection bias but does not guarantee an exactly representative result in every small sample.

Sampling variation is the difference among results that arise because different members are selected. Bias is a systematic tendency created by the selection or measurement process to favor particular results. These are related concerns but not the same thing. Increasing a properly selected sample can often make estimates more stable, while increasing a biased sample may preserve the systematic problem.

A voluntary response can be biased when people with strong opinions are more likely to respond. A survey posted only at a sports club may not represent all students' after-school preferences. A question asked in a leading way can also affect answers even if the selected people are appropriate. Evaluate who can participate, who is missing and how the question or measurement is carried out.

Representativeness depends on relevant features. If the question concerns book thickness, sampling only a shelf of oversized atlases may not represent the library's full range. If the question concerns that atlas shelf itself, the same objects may be the correct population. State the intended group and the selection frame before criticizing or defending a sample.

An informal conclusion should reflect these limits. “In this four-card sample, two were orange” is directly supported. “Half of all twenty cards must be orange” is false in the visible model. When the full population is unknown, report the sample result, explain the selection method and acknowledge that an estimate may differ from the population value. A confident tone does not make a poorly selected sample representative.

Worked population decision: A class asks about the thickness of every book on Shelf A on Friday. If the shelf holds more than ten books, measuring ten randomly selected ones gives a sample of the defined population; if it holds ten or fewer, measuring every book on it is a census. It does not automatically support a conclusion about all school-library books or next month's shelf contents.

Worked bias: Selecting only the first four model cards gives an orange share of four out of four. Since the first positions are all orange by arrangement, the method overrepresents orange cards. Randomly renaming those four cards after selecting them would not repair the original selection rule.

Worked variation: The possible samples with one and two orange cards yield 25% and 50%. Their difference shows why one small sample should not be treated as an exact population count. The sample matching the true proportion is an illustrative possible result, not a guarantee that randomness always produces perfect balance.

Sampling investigation: Create the twenty-card model and record the known population share. Draw several four-card samples using a fair number-slip procedure, returning the full population between samples but not replacing cards within one sample. Record each orange share and compare the variation. Then compare with repeated first-four convenience samples. Describe the difference between variability and systematic selection.

Independent performance: Propose a sample for an actual classroom question. Define the population, explain the selection frame, identify who or what could be omitted and describe a practical way to reduce bias. Evaluate two flawed alternatives. Report a conclusion supported by your sample and a broader claim it cannot establish. A partner checks whether your population and intended generalization refer to the same group.