返回提示词目录第 1396 页

可检验研究假设与验证方案编写

根据研究问题与设计明确变量,编写成对假设或研究目标,并列出检验方法及反证条件。

完整提示词

<context>
A hypothesis is testable when it names measurable variables, a population, a relationship with a direction or form, and an outcome that would count against it. Novice hypotheses fail in predictable ways: they restate the question ("social media affects wellbeing"), use words that cannot be measured ("better", "impacts"), pack two predictions into one, predict a causal effect from a design that can only show association, or have no result that could contradict them. Directional hypotheses need a reason from theory or prior evidence; without one, a non-directional hypothesis is more honest. Interaction and mediation hypotheses must say exactly which effect changes and how. Exploratory and qualitative questions usually should not have hypotheses at all; they get aims.
</context>

<task>
Write testable hypotheses for:
<questions>
{{questions}}
</questions>
{{#design}}
Design: {{design}}
{{/design}}

1. Classify each question as confirmatory, exploratory or qualitative. For exploratory and qualitative ones, write an aim instead and say why a hypothesis would be inappropriate.
2. For each confirmatory question, name the independent, dependent and any moderator, mediator or control variables, and the population.
3. Write each hypothesis as one prediction: the alternative (H1) with direction or form, and the null (H0). Split compound predictions.
4. Give the rationale for any direction in one line, citing the user's stated theory or evidence, or mark it as an assumption to support from the literature.
5. Match the verb to the design: "causes" or "increases" only for randomised or strong quasi-experimental designs; "is associated with" or "predicts" otherwise.
6. For each hypothesis, state the comparison or model that tests it, the result that would support it, the result that would count against it, and whether the smallest effect worth caring about has been defined.
7. Number the hypotheses so they can be carried into a preregistration and analysis plan.
</task>

<constraints>
- Every hypothesis must name measurable variables; replace vague words with what will actually be measured, or flag that the measure is undecided.
- Do not invent prior findings or theories to justify a direction.
- Do not write hypotheses for qualitative or purely descriptive questions.
- Flag any hypothesis that would need far more data than the design suggests (for example a three-way interaction).
</constraints>

<output_format>
## Question review
A table: question | type | issue found | fix.
## Hypotheses
Numbered H1/H0 pairs with variables named, and aims for non-confirmatory questions.
## How each is tested
A table: hypothesis | comparison or model | supports if | counts against if | smallest effect of interest.
## Weak spots
Bullets on causal language, missing measures, power and multiple testing.
</output_format>

可替换内容

  • {{questions}}
  • {{#design}}
  • {{design}}
  • {{/design}}

分类与标签

学习教育、研究假设、变量定义、研究设计、预注册、统计检验

前往 AI 图片创作