research-methodology

Compare original and translation side by side

🇺🇸

Original

English

🇨🇳

Translation

Chinese

Research Methodology

研究方法论

Part of Agent Skills™ by googleadsagent.ai™

属于Agent Skills™，由googleadsagent.ai™开发

Description

描述

Research Methodology guides the agent through the complete scientific research lifecycle: hypothesis generation from literature gaps, experimental design with proper controls, systematic literature review, data collection protocols, and peer review preparation. The agent functions as a research collaborator that enforces methodological rigor at every stage.

Weak methodology invalidates results regardless of how sophisticated the analysis is. This skill prevents common methodological failures: hypotheses that are unfalsifiable, experiments without proper controls, sample sizes without power analysis, and conclusions that overreach the data. The agent asks the hard questions early—before time and resources are committed to a flawed design.

The skill also covers the practical aspects of research preparation: structuring literature searches with Boolean queries, maintaining citation databases, designing reproducible experimental protocols, preparing materials for ethical review boards, and formatting submissions to meet journal-specific requirements. It bridges the gap between knowing what good research looks like and executing it systematically.

研究方法论技能可引导Agent完成完整的科研生命周期：从文献缺口生成假设、设计带有恰当对照的实验、系统性文献综述、制定数据收集方案，以及准备同行评审材料。该Agent可作为科研协作伙伴，在每个阶段确保研究方法的严谨性。

无论分析方法多么复杂，薄弱的研究方法都会导致结果无效。本技能可避免常见的方法学失误：无法证伪的假设、缺乏恰当对照的实验、未做功效分析的样本量，以及超出数据支撑范围的结论。Agent会在早期提出关键问题——在时间和资源投入到有缺陷的设计之前。

本技能还涵盖科研准备的实操环节：使用布尔查询构建文献检索策略、维护引文数据库、设计可复现的实验方案、准备伦理审查委员会所需材料，以及按照期刊特定要求格式化投稿内容。它填补了“了解优质科研标准”与“系统执行科研流程”之间的差距。

Use When

使用场景

Formulating research questions and hypotheses
Designing experiments with proper controls and sample sizes
Conducting systematic literature reviews
Preparing manuscripts for peer review submission
Writing grant proposals or research proposals
Evaluating the methodology of existing papers

制定研究问题与假设
设计带有恰当对照和合理样本量的实验
开展系统性文献综述
准备用于同行评审投稿的手稿
撰写基金申请或研究提案
评估已有论文的研究方法

How It Works

工作流程

mermaid

graph TD
    A[Research Question] --> B[Literature Review]
    B --> C[Identify Knowledge Gap]
    C --> D[Formulate Hypothesis]
    D --> E[Design Experiment]
    E --> F[Power Analysis: Sample Size]
    F --> G[Define Controls + Variables]
    G --> H[Ethical Review Preparation]
    H --> I[Data Collection Protocol]
    I --> J[Analysis Plan: Pre-registered]
    J --> K[Execute + Collect Data]
    K --> L[Analyze per Pre-registered Plan]
    L --> M[Write Manuscript]
    M --> N[Peer Review Preparation]

The methodology flow is sequential with gates: the hypothesis must be falsifiable before designing the experiment, the experiment must have adequate power before collecting data, and the analysis plan must be pre-registered before execution begins.

mermaid

graph TD
    A[Research Question] --> B[Literature Review]
    B --> C[Identify Knowledge Gap]
    C --> D[Formulate Hypothesis]
    D --> E[Design Experiment]
    E --> F[Power Analysis: Sample Size]
    F --> G[Define Controls + Variables]
    G --> H[Ethical Review Preparation]
    H --> I[Data Collection Protocol]
    I --> J[Analysis Plan: Pre-registered]
    J --> K[Execute + Collect Data]
    K --> L[Analyze per Pre-registered Plan]
    L --> M[Write Manuscript]
    M --> N[Peer Review Preparation]

该方法学流程为顺序式并设有关卡：假设必须可证伪才能进入实验设计阶段；实验必须具备足够功效才能开始数据收集；分析方案必须在执行前预先注册。

Implementation

实现代码

python

class ResearchProtocol:
    def formulate_hypothesis(self, observation: str, literature: list[str]) -> dict:
        return {
            "null_hypothesis": "There is no significant difference between...",
            "alternative_hypothesis": "Treatment X increases Y by at least Z...",
            "falsifiability": "This hypothesis is falsifiable because...",
            "variables": {
                "independent": ["Treatment type (A vs B vs control)"],
                "dependent": ["Measured outcome Y"],
                "controlled": ["Age, sex, baseline Z"],
                "confounding": ["Prior exposure to X"],
            },
        }

    def power_analysis(self, effect_size: float, alpha: float = 0.05, power: float = 0.80) -> dict:
        from statsmodels.stats.power import TTestIndPower
        analysis = TTestIndPower()
        n = analysis.solve_power(effect_size=effect_size, alpha=alpha, power=power)
        return {
            "required_n_per_group": int(np.ceil(n)),
            "total_n": int(np.ceil(n)) * 2,
            "effect_size": effect_size,
            "alpha": alpha,
            "power": power,
            "recommendation": f"Recruit {int(np.ceil(n * 1.2))} per group to account for 20% attrition",
        }

    def literature_search(self, topic: str) -> dict:
        return {
            "databases": ["PubMed", "Scopus", "Web of Science"],
            "query": f'("{topic}") AND (randomized OR controlled) AND ("2020"[Date] : "2026"[Date])',
            "inclusion_criteria": [
                "Peer-reviewed original research",
                "English language",
                "Human subjects",
                "Published 2020-2026",
            ],
            "exclusion_criteria": [
                "Review articles (captured separately)",
                "Case reports (n < 10)",
                "Non-peer-reviewed preprints",
            ],
            "prisma_flow": "Record screening, eligibility, and inclusion counts per PRISMA 2020",
        }

    def experimental_design(self, hypothesis: dict) -> dict:
        return {
            "design": "Randomized controlled trial, double-blind, parallel group",
            "randomization": "Block randomization with variable block sizes (4, 6, 8)",
            "blinding": "Participants and assessors blinded; unblinded statistician",
            "primary_outcome": hypothesis["variables"]["dependent"][0],
            "secondary_outcomes": [],
            "timeline": "Baseline → 4 weeks intervention → 8 weeks follow-up",
            "analysis_plan": "Pre-registered at OSF.io before data collection",
        }