SOC613 Chap.1 Research Questions and Causal Claims
Research Questions and Causal Claims
Quantitative analysis begins by deciding what kind of claim the evidence must support. This chapter separates descriptive, explanatory, predictive and causal questions, then connects each to a suitable comparison and conclusion. It identifies the unit of analysis, dependent variable and independent variable before calculation.
The causal section uses association, temporal order and alternative explanations as a three-part audit, with confounding as the central threat. Random sampling and random assignment are kept distinct so population representation is not mistaken for causal identification. A question about what exists needs a defensible sample and summary.
A question about how variables relate needs a comparison whose direction and magnitude can be described. Prediction adds an explicit concern with performance on outcomes not used to construct the model. Causation goes further by asking what would change under an intervention, which requires a credible counterfactual comparison. The same observed correlation can therefore answer one question well and another poorly.
The chapter develops a practical reading routine: underline the research verb, name the entity represented by each row, identify the outcome, and draw an arrow from each plausible common cause to both the explanatory variable and the outcome. This makes hidden design assumptions visible before a model gives them a numerical appearance. It also clarifies external and internal validity.
Random sampling primarily helps population representation, while random assignment primarily helps the treatment contrast. Attrition, nonresponse and noncompliance can weaken those advantages, so the final conclusion must mention the actual process rather than merely calling a study random.
By the end, a student should be able to rewrite an over-strong causal sentence as an exact association claim and state what additional evidence would justify stronger language.
What this chapter covers
- 01
Descriptive, explanatory, predictive and causal questions
- 02
Units of observation and analysis
- 03
Dependent and independent variables
- 04
Hypotheses and observable implications
- 05
Association and temporal order
- 06
Confounding and counterfactual reasoning
Worked example · free
Limit a causal claim
- 1Classify the observed relationship as an association.
- 1Name a plausible confounder such as prior motivation.
- 1Explain how it can affect both variables.
- 1Write a conclusion whose verb matches the design.
Key terms
- Unit of Analysis
- The entity about which the final conclusion is made.
- Hypothesis
- A testable expectation about a relationship between observable variables.
- Confounder
- A common cause that can make two variables appear causally related.
- Random Sampling
- Selecting units so a sample can represent a target population.
- Random Assignment
- Allocating studied units to conditions by chance to support causal comparison.
Research Questions and Causal Claims FAQ
How do I distinguish prediction from causation?
Prediction asks whether available information anticipates an outcome, while causation asks whether changing one factor would change that outcome. A variable can predict well without being a cause. State which question is being answered before choosing the method or interpretation.
Why is temporal order not enough for causation?
A proposed cause must precede its outcome, but another earlier variable may influence both. That confounder can generate the observed association. Causal reasoning therefore also needs a credible account of alternative explanations through design, adjustment and explicit assumptions.
What is the safest conclusion from an observational association?
State that the variables are associated in the studied sample, describe the direction and magnitude, and name relevant adjustment or remaining confounding. Avoid causal verbs unless the design and assumptions provide a credible comparison for the unobserved counterfactual outcome.
What does the unit of analysis change?
It determines the entity described by every statistic and conclusion. A row may represent a person, household, school or country-year. Misidentifying that entity can turn a correct calculation into a claim about the wrong social object.
Why separate random sampling from random assignment?
Sampling by chance primarily supports representation of a target population. Assignment by chance primarily supports a causal comparison within the study. A project can use one without the other, so the inference must name which mechanism actually occurred.
Exam move
For each example, write four lines before touching the data: the research question, unit of analysis, outcome and explanatory variable. Circle the main verb and classify it as descriptive, explanatory, predictive or causal. If it is causal, list one plausible common cause and state how the design addresses it. Then draft the final conclusion before calculating.
After seeing the result, revise only the value and uncertainty, not the strength of the verb. This sequence prevents a significant output from silently upgrading an observational design. Practise the distinction between random sampling and random assignment with your own scenarios, because each supports a different kind of generalisation.
Working through Research Questions and Causal Claims in SOC613? Sia is AskSia’s AI Statistics tutor — ask any SOC613 Research Questions and Causal Claims question and get a clear, step-by-step explanation grounded in how SOC613 is taught and assessed. Read this chapter free, then take your hardest questions to Sia.
