Physical Address
304 North Cardinal St.
Dorchester Center, MA 02124
D5791-23 – Standard Test Method Technical Guide
The ASTM D5791-23 standard serves as the authoritative guide for applying probability sampling methods in indoor air quality (IAQ) studies. This framework is critical for ensuring that measurements and observations taken from a subset of a population (e.g., rooms, occupants) can be mathematically projected to the entire target population with a quantifiable level of confidence. The guide provides the criteria for determining when these methods are strictly necessary for a scientific study.
🎯 Scope and Purpose of the Guide
This guide specifically covers the selection of locations for environmental monitoring equipment and the selection of building occupants for questionnaire administration. It mandates that probability sampling—where units are selected with known, non-zero probabilities—is used when the objective is to make statistically defensible inferences. Section 1.4 explicitly excludes non-probability approaches (worst-case, screening) from the scope of inference, though it acknowledges their value as informative inputs for guiding the design of a formal probability sampling plan.
✅ Core Principle of Inference: The strength of this standard lies in the concept of the Expected Value. By using a proper probability sampling design, the average result from all possible samples that could have been selected converges on a true population parameter, allowing for unbiased estimates of IAQ metrics.
⚠️ Key Distinction for Practitioners: D5791-23 is a Guide, not a Practice. For complex, multistage survey sampling procedures, the standard directs users to the references in Appendix X1. This guide establishes the architectural principles, while those references provide the detailed construction blueprints for complex designs.
🧩 Key Terminology and Concepts
Building upon Terminology D1356, this standard introduces specific sampling terms that are fundamental to understanding the implementation of IAQ studies. A clear grasp of these definitions is essential for compliance and defensibility.
| 📐 Term | 📖 Definition from D5791-23 | 💡 Application in IAQ Studies |
|---|---|---|
| Census | Survey of all elements of the target population. | Measuring formaldehyde levels in every single office on a critical floor. |
| Cluster Sample | Sampling frame partitioned into disjoint subsets (clusters); a sample of the clusters is selected. | Randomly selecting specific wings of a hospital and deploying samplers to all rooms in those wings. |
| Multistage Sample | Larger units selected at first stage; smaller units subsequently selected within those units. | Selecting city blocks, then specific buildings, then individual apartments for radon testing. |
| Compositing Samples | Physically combining the material collected in two or more environmental samples. | Combining dust wipes from several rooms into one analysis to estimate the mean lead loading. |
| Expected Value | The average value of a sample statistic over all possible samples. | The theoretical average of the mean VOC concentration across all potential sampling designs. |
⚙️ Practical Implementation for IAQ Studies
Implementation begins by matching the study objective to the sampling method. If the goal is to characterize an entire building population, the guide prescribes a shift from judgmental selection to a randomized, probability-based approach.
| 🎯 Sampling Attribute | 🟦 Probability Sampling (D5791 Focus) | ⚡ Non-Probability Sampling |
|---|---|---|
| Primary Goal | Characterize the population distribution | Identify hotspots or worst-case scenarios |
| Selection Basis | Random with known probabilities | Expert judgment or convenience |
| Basis of Inference | Statistically defensible (uses Expected Value) | Specific to selected units only |
| Common Techniques | Stratified, Cluster, Multistage, Census | Worst-case, Range Finding, Screening |
The guide is particularly valuable for complex investigations where a mixed-methods approach is warranted. Initial screening (non-probability) may identify target contaminants, which then allows for a rigorous stratified random sample (probability) to estimate the mean exposure and its variance across the building. The standard advises careful selection of the sampling frame to ensure every unit has a positive probability of selection to maintain statistical validity.
❓ Frequently Asked Questions
🔍 When is a Census required instead of a Sample per D5791-23?
A census, defined in the standard as a survey of all elements, is typically required for small, finite populations or where regulatory action demands complete enumeration. For most large-scale IAQ investigations, a carefully designed probability sample (e.g., cluster or stratified sample) provides statistically defensible results with significantly less cost and effort.
💡 How does “Compositing Samples” interact with Probability Sampling?
Compositing—physically combining materials from two or more samples—is a valid technique often used with probability sampling to estimate the population’s Expected Value while reducing analytical costs. However, the standard implies that careful design is needed, as compositing masks the variability between individual sample units, which may be required for other statistical objectives.
⚡ What is the primary advantage of a Multistage Sample design in a building?
Multistage sampling is highly advantageous when a complete list of all sampling units (frame) is unavailable or too expensive to compile. By selecting larger primary units (e.g., floors) at the first stage and then drawing smaller units (e.g., rooms) within them, the researcher focuses resources on accessible groups while maintaining the probabilistic foundation required for unbiased inference.
📌 Can I rely on “Worst-Case” Sampling instead of the methods in this guide?
No, not if you intend to generalize your findings. Section 1.4 explicitly states the standard “does not address non-probability sampling approaches.” While worst-case and screening sampling are extremely useful for developing hypotheses and guiding the design of a probability sampling plan, they lack the statistical foundation needed to make mathematically defensible inferences about the entire building population.