PES6251 Chap.3 Descriptive and Cross-sectional Studies
Descriptive and Cross-sectional Studies
Interpret the snapshot
Descriptive studies record the distribution and characteristics of a population without manipulation. Cross-sectional studies measure exposure and outcome in one observation window, making them efficient for prevalence and association but weak on temporal direction. A confounder must relate to exposure and outcome without lying on their causal pathway.
Selection, information error and confounding are different threats and require different repairs.
Protect the prevalence denominator
A percentage means little without a defined population, time and route into the sample. Report the number with the characteristic over the number actually observed, then explain whether those participants represent the target group.
A voluntary survey can estimate prevalence among respondents while still failing as an estimate for all coaches in a league. Sampling-frame gaps, differential response and missing items each narrow the denominator in a different way, so recruitment flow belongs beside the estimate.
Draw the rival explanations
Place exposure and outcome on a page, then add every plausible third variable.
Mark whether each third variable precedes exposure, follows it or sits on the pathway between exposure and outcome. Only the first category may contain confounders. This diagram prevents automatic adjustment for mediators and makes reverse causation visible.
In a one-wave survey, both directions may fit the observed pattern, and a confident causal story does not remove that structural ambiguity.
Turn each limitation into an upgrade
Repeated measurement can establish order, a broader frame can support population description, validated instruments can reduce information error and planned adjustment can address measured confounding. None solves every problem.
Choose the smallest change that addresses the claim the project actually needs. This is stronger than ending with correlation is not causation, because it states why the current evidence stops and what the next design must contribute.
What this chapter covers
- 01
Prevalence and its denominator
- 02
Cross-sectional time structure
- 03
Reverse causation
- 04
Three conditions for confounding
- 05
Selection and information bias
- 06
Design upgrades for unresolved uncertainty
Worked example · free
Audit a one-wave coaching survey
- 2State what was measured directly.
- 4Name temporal and confounding alternatives.
- 3Write a defensible conclusion and next design.
Key terms
- Prevalence
- The proportion of a defined population with a characteristic during a stated period or at a stated time.
- Reverse Causation
- An explanation in which the apparent outcome actually influences the presumed exposure.
- Confounder
- A third variable related to exposure and outcome that is not part of the causal pathway.
- Selection Bias
- Distortion arising when entry or retention relates to variables central to the study.
- Information Bias
- Systematic error arising from inaccurate or unequal measurement of exposure, outcome or other variables.
Descriptive and Cross-sectional Studies FAQ
Why can a cross-sectional study not usually establish direction?
Exposure and outcome are measured in the same observation window, so the design cannot usually show which came first. The observed association may reflect exposure influencing outcome, outcome influencing exposure or a third factor influencing both. Repeated measurement or prospective follow-up is needed when temporal sequence is central.
Is every third variable a confounder?
No. A confounder must be associated with the exposure, associated with the outcome and outside the causal path between them. A variable caused by the exposure and helping produce the outcome is a mediator. Automatically controlling mediators can remove part of the effect the study intends to estimate.
How should prevalence be reported from a voluntary survey?
Name the respondent denominator and avoid treating respondents as the entire target population without evidence. Report who was eligible, invited and completed the survey, then assess whether response could relate to the characteristic measured. A random invitation does not preserve representativeness when non-response is strongly selective.
Assessment move
Draw exposure, outcome and each candidate third variable. Mark whether each third variable precedes exposure or sits on the pathway. For every cross-sectional result, write one association sentence and one uncertainty sentence, then propose the smallest design change that would resolve the uncertainty you care about most. Audit every percentage by writing its numerator, denominator, target population and observation window.
Next, trace recruitment from the sampling frame to respondents and predict who may be missing. Separate selection error, measurement error and confounding because each needs a different repair. Finally, replace causal verbs with language the single observation window can support and specify whether repeated measurement, better sampling or planned adjustment is the priority.
Working through Descriptive and Cross-sectional Studies in PES6251? Sia is AskSia’s AI Statistics tutor — ask any PES6251 Descriptive and Cross-sectional Studies question and get a clear, step-by-step explanation grounded in how PES6251 is taught and assessed. Read this chapter free, then take your hardest questions to Sia.
