Why Parameterization Starts With the Population
A hypergeometric count describes how many designated members of a fixed population appear in a sample drawn without replacement. The main challenge in a word problem is often not calculation: it is deciding exactly which units belong to the population, which of them count as successes, and how many units are sampled. A misplaced population boundary or a mistaken sample size changes the model before any probability is considered.
The word success is a label for one of two categories in the model. It does not mean that the outcome is desirable. A defective component, an incomplete record, or a policy with a specified feature can all be designated as a success if that is the event being counted.
These roles are not interchangeable. \(N\) counts all eligible units before the draw; \(K\) counts the eligible units marked as successes; \(n\) counts the units selected. By contrast, a value such as \(x=3\) is a possible observed count of successes in the sample, not a model parameter.
Define the Model Before Assigning the Numbers
A reliable parameterization begins with the sampling design, not with the first number mentioned in the problem. State the population boundary and the unit being sampled. Then identify the characteristic to be counted and tally how many eligible units have it. Finally, identify how many distinct units are selected.
Specify which distinct units could be selected at the start of the draw. Their total is \(N\).
State the characteristic to be counted. Count the population units with that characteristic; this total is \(K\).
Count how many units are selected. For one hypergeometric count, the sample must be selected without replacement, with all size-\(n\) subsets equally likely.
Define \(X\) as the number of sampled units with the designated characteristic. Then write \(X\sim\operatorname{Hypergeometric}(N,K,n)\).
The model assumes that the population and its success labels are fixed during sampling, that units are selected without replacement, and that the sample is a simple random sample of size \(n\). The last condition means every subset of \(n\) population units is equally likely. Without random selection, the count may still be recorded, but the hypergeometric model is not automatically justified.
Why These Three Quantities Determine the Count Model
The parameter meanings follow directly from the setup. Imagine marking each of the \(N\) eligible units either success or failure before any draw. There are \(K\) marked successes and \(N-K\) failures. Selecting \(n\) units without replacement produces a sample containing some number \(X\) of the success-marked units. Thus the total number of available units, the number marked success, and the number drawn are precisely the information required to describe this count.
The simple-random-sample condition matters because it prevents the selection procedure from favoring particular units or subsets. For example, drawing \(n\) units by taking all units from one category or by deliberately choosing a fixed number from each subgroup is not generally the same sampling design as choosing one size-\(n\) subset uniformly from the whole population.
This identification does not require knowing the eventual value of \(X\). Parameters are set by the population and the sampling plan before the draw. Once the sample is observed, its success count is data. Keeping those two stages separate prevents a common circular error: using an observed sample count as though it were the number of successes in the whole population.
Worked Parameterizations
Worked Example: Inspection of a Component Lot
A hypothetical lot contains 96 distinct valve cartridges, of which 12 fail a specified pressure check. An inspector selects 15 cartridges uniformly at random without replacement. Let \(X\) be the number in the sample that fail. Identify the hypergeometric parameters and translate the event “at least four failures” into notation.
Method. First identify the complete set from which the inspector draws: all 96 cartridges in the lot. Therefore \(N=96\). The designated success is “fails the pressure check,” and the lot contains 12 such cartridges, so \(K=12\). The inspector draws 15 distinct cartridges, so \(n=15\).
The event “at least four failures” is \(X\geq 4\). The number 4 describes the event threshold; it does not replace \(n=15\), the sample size. Nor does the sample size become 4 just because the question names four failures.
Check. The population has \(96-12=84\) cartridges that do not fail, and \(12+84=96\). The requested sample has 15 cartridges, distinct and selected without replacement, matching the model assumptions. No probability calculation is needed to identify the parameters.
Common mistake avoided: setting \(n=4\) from “at least four.” That confuses a possible value of the count with the number of units drawn; the correct sample size is 15.
Worked Example: Defining Success as a Complete File
A hypothetical archive has 240 eligible claim files. Of these, 54 lack a signed authorization form. An auditor randomly selects 20 files without replacement. Let \(Y\) count files in the sample that have a signed authorization form. Identify the parameters for \(Y\), and state how the count would be parameterized if the auditor instead counted files that lack the form.
Method. The population is the 240 eligible files, so \(N=240\). For \(Y\), success means “has a signed authorization form.” The number of population successes is the number of files not in the stated missing-form category:
The auditor selects 20 files, so \(n=20\). Therefore,
If instead \(Z\) counts files that lack the form, the population and sampling plan do not change: \(N=240\) and \(n=20\). Only the designated success changes. There are 54 such files, giving
Check. The two success categories partition the eligible files: \(186+54=240\). Both models use a sample size of 20. The parameter \(K\) changes because the question changes what counts as a success; the word itself has no fixed positive or negative meaning.
Common mistake avoided: using \(K=54\) for \(Y\) simply because 54 is the number highlighted in the description. That would model the count of files lacking the form, not the count of files with it.
Worked Example: A Sampling Plan That Is Not One Simple Random Sample
A hypothetical review covers 180 policies from two divisions. Division A has 100 policies, including 20 with a specified rider. Division B has 80 policies, including 16 with the rider. A review team selects 10 policies at random from Division A and 10 at random from Division B, each without replacement within its division. A candidate proposes a single hypergeometric model with \(N=180\), \(K=36\), and \(n=20\) for the total number of reviewed policies with the rider. Is that parameterization justified by the stated sampling plan?
Method. The proposed population total and rider count are tallied correctly:
However, correct totals alone do not establish a single hypergeometric model. The design fixes the number selected from each division. It does not choose one size-20 subset uniformly from all 180 policies: every selected sample must contain exactly 10 policies from A and exactly 10 from B. Therefore the proposed single model with parameters \((180,36,20)\) is not justified by this design.
One can define \(X_A\) as the number with the rider among the 10 selected from A, and \(X_B\) as the number with the rider among the 10 selected from B. Their individual parameterizations are
The total count is \(X_A+X_B\), but it is generated by two division-specific draws. The parameters for each draw reflect its own eligible population and success count.
Check. The two sample sizes add to the total reviewed count, \(10+10=20\), and the population counts add to \(180\). Those arithmetic checks are necessary but not sufficient: the stipulated division quotas also matter.
Common mistake avoided: assuming that \(N\), \(K\), and \(n\) totals always imply a simple random sample from the combined population. The selection design must also match that model.
Common Parameterization Errors
- Using the sample count as \(K\). \(K\) counts successes in the entire eligible population. An observed sample count is a value of \(X\).
- Using the event threshold as \(n\). If a sample of 15 is inspected and the question asks about at least 4 successes, then \(n=15\); 4 is a threshold for \(X\).
- Leaving the population boundary vague. A lot, a set of remaining records, and all records originally held by an office are different populations if eligibility differs.
- Assuming “success” means a favorable result. It means only the category being counted. Define it explicitly in context.
- Ignoring the design. Sampling with replacement, nonrandom selection, or fixed subgroup quotas does not automatically produce one hypergeometric count for the combined population.
Check Your Understanding
For each problem, define the count and identify the parameters. Check that the eligible population and sampling design are explicit.
- A hypothetical crate contains 75 sealed modules, including 9 with a damaged label. A random sample of 8 is drawn without replacement. Let \(X\) count damaged labels. State the model parameters. (Answer: \(X\sim\operatorname{Hypergeometric}(75,9,8)\).)
- A hypothetical database has 320 eligible accounts, 48 of which have an inactive status. A random sample of 25 is taken without replacement. Let \(Y\) count active accounts. Identify the number of population successes and the model for \(Y\). (Answer: \(K=320-48=272\), so \(Y\sim\operatorname{Hypergeometric}(320,272,25)\).)
- A hypothetical warehouse has 140 eligible packages, of which 35 require temperature-controlled handling. A random sample of 12 packages is inspected without replacement. If \(X\) counts such packages, translate “fewer than three” into an event and give the parameters. (Answer: \(X\sim\operatorname{Hypergeometric}(140,35,12)\); the event is \(X<3\).)
- A hypothetical portfolio has 90 eligible contracts, 18 with an optional benefit. A simple random sample of 14 is taken without replacement. After selection, the reviewer finds 4 contracts with the benefit. Distinguish the model parameters from the observed count. (Answer: \(N=90\), \(K=18\), \(n=14\); the observed count is \(x=4\), not a parameter.)
- A hypothetical review has 60 records in Group A, 12 flagged, and 40 records in Group B, 10 flagged. The reviewer selects 6 from A and 4 from B, randomly within each group and without replacement. Identify the group-specific models and explain whether one simple random sample of size 10 from the combined 100 records is described. (Answer: \(X_A\sim\operatorname{Hypergeometric}(60,12,6)\) and \(X_B\sim\operatorname{Hypergeometric}(40,10,4)\); the fixed group quotas do not describe one simple random sample of 10 from all 100.)