POLS1009 Chap.8 Writing and Testing Hypotheses
Writing and Testing Hypotheses
This chapter covers the Week 6 research strand, where a theory becomes something you can test. A hypothesis is a conjecture about how an outcome relates to an explanation, written in the course notation as a signed dependent variable on the left and a rising independent variable on the right, and phrased in indicators rather than concepts.
You will see how a deductive hypothesis is reasoned from theory while an inductive one comes from patterns in data, and what three things every usable hypothesis must do. A path diagram then sorts every variable by role, from explanation and outcome to mediators, moderators, controls and confounders, and shows why a role belongs to a study rather than to a variable.
The middle of the chapter works through the four criteria and the six common errors, repairing a broken statement for each, and notes that the Week 6 review groups those errors in a different way. It ends with the null and research hypotheses, the logic of rejecting the null, the threats that remain after a significant result, and what to check before blaming the theory.
What this chapter covers
- 01
What a hypothesis is and how the course notation writes it
- 02
Indicators, level of analysis and probabilistic claims
- 03
Deductive and inductive routes to a hypothesis
- 04
Six variable roles, including mediators, moderators and confounders
- 05
Four criteria and the one-sign rule
- 06
Six common errors and how to repair each
- 07
Null versus research hypotheses and rejecting the null
- 08
Threats to causal inference and a pre-test checklist
Worked example · free
Assigning variable roles in a path diagram for a community grants study
- 1IV, DV and hypothesis. The IV is grant dollars per resident and the DV is volunteer hours per resident, measured by suburb: + Volunteer hours ← + Grant dollars.
- 1Mediator. The number of local events lies on the path between grants and volunteering and carries the mechanism, since events are where recruitment happens.
- 1Moderator. The share of residents working weekends changes the strength of the grants to volunteering link, so its arrow points at the link, not at either variable.
- 1Confounder. Prior association size influences both who receives grants and how much volunteering already happens, so it could make the relationship spurious.
- 1Control. The retired share of residents is held constant, or adjusted for statistically, so the comparison isolates the grants' own effect; prior association size, the confounder, must also be adjusted for in the design.
Key terms
- Independent variable
- The factor a hypothesis offers as the explanation, whose changing values are expected to produce changes in the outcome being studied.
- Dependent variable
- The outcome a study sets out to explain, named for the expectation that its values depend on those of the explanatory variable.
- Mediating variable
- A variable that sits on the path from explanation to outcome and carries the mechanism, so the effect of the explanation travels through it.
- Moderating variable
- A variable that changes how strong the link between explanation and outcome is, or which way it runs, without lying on the causal path itself.
- Alternative hypothesis
- The research claim that two variables are related, usually in a stated direction, which is supported only when the evidence allows the null of no relationship to be rejected.
Writing and Testing Hypotheses FAQ
Which criteria must a well-formed hypothesis meet?
The Week 6 slides list four: it contains no proper names, it is completely specified, it is testable and it is falsifiable. It should also link an outcome to a proposed explanation, say in which way the two are connected, and make clear what a test would look for. A hypothesis that puts both signs in front of the dependent variable fails, because any result would count as support.
How do I tell a mediator from a moderator?
Ask whether the variable explains how the effect happens or when it is stronger. A mediator is the step in between: the cause changes it, and it changes the outcome. A moderator stays off the path and alters the size or sign of the effect, for example making a policy work in some places but barely at all in others.
Why can a research hypothesis never say there is no relationship?
Because that statement is reserved for the null, which is the benchmark the test tries to reject. Evidence can count against a claim of no relationship, but failing to find a relationship never proves its absence, since weak measures or too few cases can hide a real effect. Your own claim therefore has to predict a relationship.
What is an improper comparison in a hypothesis?
It is a comparison drawn across categories of the outcome rather than categories of the explanation. Saying one group prefers policy A more than policy B compares outcomes; a testable version compares two groups on the same outcome. Week 6 calls this the most frequent slip, and the review slide folds it together with a missing comparison.
What should I do if my results fail to reject the null?
Do not conclude straight away that the theory is wrong. First ask whether the measures captured the concepts, whether the sample was large enough to detect an effect, whether the cases fall within the theory's scope, and whether the model was specified sensibly. Report effect sizes with confidence intervals so readers can judge how much the evidence actually rules out.
Exam move
Practise on sentences you meet in news reports and readings. For each causal claim, write it in the course notation, then mark which variable is the explanation and which the outcome, and add the mediator the author implies. Next, run the claim through the four criteria and say which of the six errors it commits, if any, and rewrite it.
Keep a short personal list of broken and repaired statements on topics of your own, because inventing the repair is what builds the skill. Finally, for every hypothesis you write, also write its null and one threat to inference, so that the testing side becomes automatic rather than something you remember at the end.
Working through Writing and Testing Hypotheses in POLS1009? Sia is AskSia’s AI Arts and Humanities tutor — ask any POLS1009 Writing and Testing Hypotheses question and get a clear, step-by-step explanation grounded in how POLS1009 is taught and assessed. Read this chapter free, then take your hardest questions to Sia.