研究方法学撰写助手

研究方法

研究设计 本研究采用混合方法的课堂行动研究取向，旨在在真实教学情境中迭代优化教学实践并检验其成效。整体设计为嵌入式混合方法：在每一轮“计划—行动—观察—反思”循环中同步收集定量与定性数据，并在循环内和循环间进行整合，用以解释学习成效与课堂过程之间的关联（Creswell & Plano Clark, 2018；Kemmis, McTaggart, & Nixon, 2014）。定量部分以前后测与标准化观察量表为主，定性部分以半结构访谈与研究者反思札记为主。研究目的兼顾改进与解释：一方面通过行动研究提升教学实践的适切性与有效性，另一方面通过多源证据建立干预成效与机制的合理解释（Greene, Caracelli, & Graham, 1989；Mertler, 2017）。

研究场域与参与者 研究在一所常态化教学班级的真实课堂开展。采用自然情境中的整群样本（intact class）并辅以目的性抽样用于访谈，以确保观点的多样性与信息丰富度（Cohen, Manion, & Morrison, 2018；Patton, 2015）。参与者包括任课教师、班级学生以及在部分课时参与的同伴观察者。为降低情境偏差，保留原有课程安排与评价制度，仅将干预策略嵌入既有教学流程。访谈对象根据学习表现、课堂参与度与性别等维度进行最大变异抽样；访谈样本规模遵循“信息力”原则，在达到信息饱和时停止招募（Malterud, Siersma, & Guassora, 2016）。

教学干预与行动周期 研究预计实施多轮行动研究循环。每轮包含：

• 计划：基于上一轮证据制定可操作的教学改进方案，明确可观察的目标与成功标准。
• 行动：在连续课次中实施干预策略，记录实施剂量（课时数、时长）、遵循度（关键步骤完成度）与质量（师生互动品质）。
• 观察：使用课堂观察量表与过程性记录收集数据，并实施必要的同伴观课。
• 反思：汇总定量与定性证据，识别有效做法与瓶颈，修订下一轮方案（Kemmis et al., 2014；Mertler, 2017）。

测量工具与资料来源

1. 课堂观察量表

• 工具构建：参照既有课堂观察工具的构念结构（如CLASS、RTOP）界定维度与操作性定义，形成4点或5点行为锚定评分量表，涵盖课堂结构、学术严谨性、学习者参与、形成性评价与师生互动等核心维度（Pianta, La Paro, & Hamre, 2008；Sawada et al., 2002）。
• 内容效度：邀请3–5位学科与测量专家进行德尔菲式评审，依据相关性、可观察性与表述清晰度修订条目（AERA, APA, & NCME, 2014；Messick, 1995）。
• 评分者培训与信度：采用编码手册进行校准训练，使用示例视频与现场双人同录。抽取至少20%课时进行双评并计算组内相关系数ICC(2,k)评估一致性，目标值≥0.75；对关键事件的类别判断计算Cohen’s κ（Shrout & Fleiss, 1979）。内部一致性以Cronbach’s α评估（Cronbach, 1951）。
• 结构检验：在样本允许时进行探索性/验证性因素分析以检验量表维度的拟合性与区分度（Cohen et al., 2018）。

2. 半结构访谈

• 访谈提纲：围绕学习体验、策略可用性、参与动机、困难与支持、课堂互动质量与公平性等主题，设置开放性主问题与追问（Patton, 2015）。
• 实施与记录：课后48小时内开展个别或小组访谈，时长依受访者负担与信息饱和度确定；经允诺后录音并转写。
• 质性严谨性：通过成员检核回馈要点与初步主题；保留审计轨迹（编码本、备忘录与决策记录）以提升可审查性（Lincoln & Guba, 1985）。

3. 前后测（学习成效）

• 工具：依据课程目标与认知层次（识记、理解、迁移/应用）编制或选用与课程一致的测验，兼顾客观题与开放性题。采用专家评审与小样本预测试检验内容效度与可理解性（AERA et al., 2014）。
• 计分与信度：客观题采用CTT指标计算KR‑20/α；主观题以评分量表双评计分，计算ICC/κ评估评分一致性。必要时采用锚题或Rasch模型对难度差异做校正（Bond & Fox, 2015）。
• 安排：前测于干预前一周内实施，后测于干预结束后同等条件下实施，确保测验条件与时间等价以控制测试效应（Shadish, Cook, & Campbell, 2002）。

4. 实施忠实度（Fidelity）

• 维度：遵循度（核心步骤完成比例）、剂量（次数/时长）、质量（教学互动质量）与差异化适配（对不同学习者的调整）。
• 证据：教师自评清单、同伴观察记录与课堂录影的清单化编码（O’Donnell, 2008）。

数据收集程序

• 基线：收集前测成绩与首轮常态课堂观察数据，建立可比基线。
• 循环执行：每轮干预期内，安排至少2次系统性课堂观察与一次短访谈/焦点访谈；必要时补充形成性测验。
• 反思与修订：每轮结束后进行多源证据汇整会议，生成可操作的教学修订点并进入下一轮。

数据分析

1. 定量分析

• 描述统计：报告各指标均值、标准差与95%置信区间；对量表维度提供内部一致性系数。
• 成效检验：对同一群体前后测采用配对样本t检验；若不满足正态性，采用Wilcoxon符号秩检验。多轮循环可采用线性混合模型检验时间趋势与循环效应，必要时控制协变量（Cohen, 1988；Field方法学见Cohen et al., 2018）。
• 效应量：报告Cohen’s d（配对校正）与相应置信区间，便于累积科学解释；对非参数检验报告r或相应转换的效应量（Lakens, 2013）。
• 观察量表：对关键维度进行前后比较或分阶段趋势分析；双评数据计算ICC/κ并报告置信区间（Shrout & Fleiss, 1979）。
• 缺失数据：描述缺失模式；若>5%且非完全随机，采用多重插补并进行敏感性分析（Rubin, 1987）。

2. 定性分析

• 编码流程：采用反复比较的主题分析，历经熟悉资料、初始编码、主题生成、主题审订与命名等阶段；编码本包含定义、纳入/排除标准与示例（Braun & Clarke, 2006）。
• 可靠性：对不低于20%的材料进行双人独立编码并计算κ或百分一致；分歧通过讨论达成一致并更新编码本。
• 解释策略：注重由数据出发的模式识别，同时对照理论范式与行动研究目标进行解释性编码，追踪机制线索（Patton, 2015）。

3. 混合整合

• 循环内汇合：采用并行汇合策略，将同一循环的量化与质性结果进行对照，构建“联结式展示”（joint display），识别趋同、互补或不一致证据（Fetters, Curry, & Creswell, 2013）。
• 循环间解释：运用“线索追踪”（following a thread）法在跨循环追踪关键主题与量化指标变化，生成元推论以指导下一轮行动（Creswell & Plano Clark, 2018）。
• 解释优先级：当结果不一致时，以实施忠实度、工具信度与数据来源层级进行权衡，并以教学情境证据为最终解释锚点（Greene et al., 1989）。

质量保障与伦理

• 测量效度与信度：遵循“统一效度观”，从内容、结构、关系与后果等方面积累效度证据；报告内部一致性、评分者一致性与（可行时的）结构指标（Messick, 1995；AERA et al., 2014）。
• 可信度与可转移性：采用数据三角互证、成员检核、同伴审议与厚描写提高可信度与可转移性；通过研究日志与决策备忘录建立审计轨迹（Lincoln & Guba, 1985）。
• 伦理：获得机构审查与家长/学生知情同意，确保自愿参与与可随时退出；采取去标识化与安全存储；访谈与观察强调最小干预，不影响常态教学。

效度威胁与控制

• 内在效度：通过前后测、对照既往基线、重复循环与忠实度监测缓解成熟、历史与测试效应威胁（Shadish et al., 2002；O’Donnell, 2008）。
• 外在效度：在自然课堂中实施并进行情境厚描，界定推论边界；报告班级特征与实施条件以便情境化迁移（Cohen et al., 2018）。

参考文献

• AERA, APA, & NCME. (2014). Standards for Educational and Psychological Testing. AERA.
• Bond, T. G., & Fox, C. M. (2015). Applying the Rasch Model: Fundamental Measurement in the Human Sciences (3rd ed.). Routledge.
• Braun, V., & Clarke, V. (2006). Using thematic analysis in psychology. Qualitative Research in Psychology, 3(2), 77–101.
• Cohen, J. (1988). Statistical Power Analysis for the Behavioral Sciences (2nd ed.). Lawrence Erlbaum.
• Cohen, L., Manion, L., & Morrison, K. (2018). Research Methods in Education (8th ed.). Routledge.
• Creswell, J. W., & Plano Clark, V. L. (2018). Designing and Conducting Mixed Methods Research (3rd ed.). Sage.
• Cronbach, L. J. (1951). Coefficient alpha and the internal structure of tests. Psychometrika, 16, 297–334.
• Fetters, M. D., Curry, L. A., & Creswell, J. W. (2013). Achieving integration in mixed methods designs—Principles and practices. Health Services Research, 48(6 Pt 2), 2134–2156.
• Greene, J. C., Caracelli, V. J., & Graham, W. F. (1989). Toward a conceptual framework for mixed-method evaluation designs. Educational Evaluation and Policy Analysis, 11(3), 255–274.
• Kemmis, S., McTaggart, R., & Nixon, R. (2014). The Action Research Planner: Doing Critical Participatory Action Research. Springer.
• Lakens, D. (2013). Calculating and reporting effect sizes to facilitate cumulative science. Frontiers in Psychology, 4, 863.
• Lincoln, Y. S., & Guba, E. G. (1985). Naturalistic Inquiry. Sage.
• Malterud, K., Siersma, V. D., & Guassora, A. D. (2016). Sample size in qualitative interview studies: Guided by information power. Qualitative Health Research, 26(13), 1753–1760.
• Mertler, C. A. (2017). Action Research: Improving Schools and Empowering Educators (5th ed.). Sage.
• O’Donnell, C. L. (2008). Defining, conceptualizing, and measuring fidelity of implementation. Review of Educational Research, 78(1), 33–84.
• Patton, M. Q. (2015). Qualitative Research & Evaluation Methods (4th ed.). Sage.
• Pianta, R. C., La Paro, K., & Hamre, B. K. (2008). Classroom Assessment Scoring System (CLASS) Manual, K–3. Paul H. Brookes.
• Sawada, D., et al. (2002). Measuring reform practices in science and mathematics classrooms: The Reformed Teaching Observation Protocol. School Science and Mathematics, 102(6), 245–253.
• Shadish, W. R., Cook, T. D., & Campbell, D. T. (2002). Experimental and Quasi-Experimental Designs for Generalized Causal Inference. Houghton Mifflin.
• Shrout, P. E., & Fleiss, J. L. (1979). Intraclass correlations: Uses in assessing rater reliability. Psychological Bulletin, 86(2), 420–428.

Research Design and Rationale This study employs an explanatory sequential mixed-methods design, administering a structured survey first and then conducting focus group interviews to explain and elaborate on quantitative patterns (Creswell & Plano Clark, 2017; Tashakkori & Teddlie, 2010). Integration is planned at design (connecting qualitative sampling to survey results), methods (building the focus group guide from quantitative findings), and interpretation (joint displays and meta-inferences) to enhance validity through triangulation and complementarity (Fetters, Curry, & Creswell, 2013).

Setting, Participants, and Sampling

• Population and frame: Educators (and, where appropriate, students) from a defined set of schools or programs. The sampling frame is derived from institutional rosters or partner districts.
• Survey sampling: Stratified random sampling by institution type and demographic strata to improve representativeness and precision (Fowler, 2014). Planned target N ≈ 400–600 respondents, which affords ≥.80 power to detect small-to-moderate effects (f2 ≈ .03–.05) in multivariable models, assuming α = .05 (Cohen, 1988). Where clustering by school/classroom is non-negligible, design effects will inform the target N and multilevel models will be used (Raudenbush & Bryk, 2002).
• Focus groups: Purposeful maximum-variation sampling of survey respondents who consented to follow-up, selected to reflect key subgroups (e.g., role, experience, context) and contrasting quantitative profiles (e.g., high/low scores on focal constructs). Approximately 6–8 groups with 6–8 participants each are planned, with final numbers determined by thematic sufficiency and the diversity of perspectives necessary to explain survey results (Morgan, 1997; Krueger & Casey, 2015).

Instruments and Measures

• Survey: A composite instrument assessing targeted constructs (e.g., instructional practices, self-efficacy, implementation barriers) using established scales where possible; new items are developed following best practices in scale development and psychometric validation (DeVellis, 2017). Items use 5-point Likert-type formats, with balanced keying and clear anchors to reduce satisficing (Krosnick, 1991).
• Instrument development: Expert review for content validity; cognitive interviewing with 8–12 participants to refine item wording, comprehension, and response processes (Willis, 2005). A pilot test (n ≈ 50–80) will evaluate reliability and factor structure.
• Focus group protocol: A semi-structured guide linked to survey findings (e.g., areas of divergence, surprising regression results, or subgroup differences) and designed to elicit explanations, contextual factors, and implementation details (Krueger & Casey, 2015). Prompts are neutral, open-ended, and ordered from general to specific to minimize priming effects (Morgan, 1997).

Data Collection Procedures

• Survey administration: Online mode with mobile-optimized design; three to four tailored contacts (pre-notice, invitation, reminder sequences) to maximize response and reduce nonresponse bias (Dillman, Smyth, & Christian, 2014). Incentives are modest and ethically appropriate. Estimated completion time is ≤15 minutes to limit burden.
• Focus groups: 60–90 minutes each, in person or via secure video-conference. A trained moderator and an assistant conduct sessions using a standardized protocol; discussions are audio-recorded and professionally transcribed. Ground rules emphasize confidentiality while clarifying its limits in group settings (Krueger & Casey, 2015).

Quantitative Data Analysis

• Preparation: Data cleaning, screening for outliers, evaluation of missingness mechanisms. Missing data will be addressed via multiple imputation under MAR assumptions, with sensitivity analyses (Rubin, 1987; Schafer & Graham, 2002).
• Measurement: Internal consistency reliability (Cronbach’s alpha and McDonald’s omega where appropriate); confirmatory factor analysis (CFA) to assess construct validity and (if applicable) measurement invariance across key subgroups prior to between-group comparisons (DeVellis, 2017).
• Modeling: Descriptive statistics with 95% confidence intervals; multivariable regression or multilevel models if clustering is present (Raudenbush & Bryk, 2002). Model diagnostics will assess linearity, multicollinearity, and residual assumptions. Multiple comparisons will be controlled using false discovery rate procedures when applicable (Benjamini & Hochberg, 1995). Effect sizes (e.g., standardized betas) will accompany p-values (Cohen, 1988).
• Bias checks: Nonresponse bias will be assessed through frame–respondent comparisons when auxiliary data are available and via late-responder analyses (Groves, 2006). Common method bias will be mitigated procedurally (assuring anonymity, psychologically separating measures, varied scale formats) and evaluated statistically (e.g., marker variable or latent method factor models) (Podsakoff et al., 2003).

Qualitative Data Analysis

• Approach: Reflexive thematic analysis conducted iteratively and systematically to identify explanatory mechanisms and contextual contingencies that account for quantitative patterns (Braun & Clarke, 2006).
• Coding: A codebook will be generated deductively from the research questions and key quantitative results and inductively from the data. Two trained coders will independently code an initial subset to refine the codebook; intercoder agreement will be examined (e.g., Cohen’s kappa) to inform training and code refinement, while final analysis emphasizes analytic rigor and transparency rather than kappa thresholds alone (Cohen, 1960; Braun & Clarke, 2006).
• Trustworthiness: Strategies include an audit trail, reflexive memos, peer debriefing, and targeted member checking of interpretive summaries where feasible (Lincoln & Guba, 1985). Reporting will follow relevant standards (e.g., COREQ) adapted to focus groups (Tong, Sainsbury, & Craig, 2007).

Mixed-Methods Integration

• Connecting: Quantitative results will inform qualitative sampling by selecting participants representing key subgroups or outlier patterns.
• Building: Focus group questions will probe unexpected or theory-relevant quantitative findings.
• Merging: Joint displays will align quantitative estimates with qualitative themes to generate meta-inferences, prioritizing convergence, complementarity, or explanation of divergence (Fetters et al., 2013; Creswell & Plano Clark, 2017).

Quality Assurance and Data Management

• Standardization: Detailed field manuals for survey administration and focus group facilitation; training and calibration of data collectors and moderators; pilot rehearsals.
• Instrument fidelity: Cognitive testing and pilot analyses; timing checks to identify satisficing; embedded attention checks with minimal disruption (Krosnick, 1991).
• Data handling: Secure storage, encryption, and role-based access; de-identification of transcripts; version-controlled analysis scripts and preregistered quantitative analysis plans to enhance transparency (where feasible).
• Ethical compliance: Institutional ethical approval; informed consent; right to withdraw without penalty; data minimization; confidentiality safeguards consistent with BERA guidelines (BERA, 2018).

Risks and Mitigation

• Low survey response rate: Tailored contact protocol, mixed-mode follow-up if needed, modest incentives, and short instrument (Dillman et al., 2014). Nonresponse weighting if auxiliary data allow (Groves, 2006).
• Scheduling and recruitment challenges for focus groups: Flexible scheduling (including virtual sessions), oversampling consenting respondents, and offering alternative small-group or paired interviews where necessary (Morgan, 1997).
• Technology failures (online survey platform or virtual focus groups): Redundant systems, pre-session tech checks, and contingency recording solutions.
• Social desirability and group conformity: Neutral phrasing, clear confidentiality norms, skilled moderation, and triangulation with survey findings to detect inconsistencies (Krueger & Casey, 2015).
• Data loss or confidentiality breach: Encrypted storage, regular backups, and minimal linking files kept separately; strict access controls.
• Unexpected contextual disruptions (e.g., policy changes, school closures): Use of remote modalities, extended data collection window, and adaptive scheduling.

Milestones and Timeline (12 months)

• Months 1–2: Ethics approval; finalize design; instrument drafting; expert review; cognitive interviews.
• Month 3: Pilot testing (survey and focus group guide); instrument revision; preregistration of quantitative analysis plan.
• Months 4–5: Main survey rollout; reminder protocol; begin preliminary cleaning.
• Months 6–7: Focus group recruitment and data collection; ongoing transcription.
• Months 7–8: Quantitative data cleaning, imputation, and measurement modeling; descriptive and preliminary inferential analyses.
• Months 8–9: Qualitative coding and thematic analysis; intercoder calibration; audit trail consolidation.
• Month 10: Integration via joint displays; development of meta-inferences.
• Month 11: Member checking of interpretive summaries; sensitivity analyses; finalize results.
• Month 12: Manuscript/report preparation and dissemination; archive de-identified data and analysis scripts as appropriate.

Limitations and Delimitations

• Focus groups limit individual confidentiality; results will be interpreted with this constraint acknowledged. Survey results may be subject to residual nonresponse or common method bias; procedural and statistical controls aim to minimize these threats.

References

• Benjamini, Y., & Hochberg, Y. (1995). Controlling the false discovery rate: A practical and powerful approach to multiple testing. Journal of the Royal Statistical Society Series B, 57(1), 289–300.
• Braun, V., & Clarke, V. (2006). Using thematic analysis in psychology. Qualitative Research in Psychology, 3(2), 77–101.
• British Educational Research Association (BERA). (2018). Ethical guidelines for educational research (4th ed.).
• Cohen, J. (1988). Statistical power analysis for the behavioral sciences (2nd ed.). Routledge.
• Cohen, J. (1960). A coefficient of agreement for nominal scales. Educational and Psychological Measurement, 20, 37–46.
• Creswell, J. W., & Plano Clark, V. L. (2017). Designing and conducting mixed methods research (3rd ed.). SAGE.
• DeVellis, R. F. (2017). Scale development: Theory and applications (4th ed.). SAGE.
• Dillman, D. A., Smyth, J. D., & Christian, L. M. (2014). Internet, phone, mail, and mixed-mode surveys: The tailored design method (4th ed.). Wiley.
• Fetters, M. D., Curry, L. A., & Creswell, J. W. (2013). Achieving integration in mixed methods designs—Principles and practices. Health Services Research, 48(6 Pt 2), 2134–2156.
• Fowler, F. J. (2014). Survey research methods (5th ed.). SAGE.
• Groves, R. M. (2006). Nonresponse rates and nonresponse bias in household surveys. Public Opinion Quarterly, 70(5), 646–675.
• Krueger, R. A., & Casey, M. A. (2015). Focus groups: A practical guide for applied research (5th ed.). SAGE.
• Krosnick, J. A. (1991). Response strategies for coping with the cognitive demands of attitude measures in surveys. Applied Cognitive Psychology, 5(3), 213–236.
• Lincoln, Y. S., & Guba, E. G. (1985). Naturalistic inquiry. SAGE.
• Morgan, D. L. (1997). Focus groups as qualitative research (2nd ed.). SAGE.
• Podsakoff, P. M., MacKenzie, S. B., Lee, J.-Y., & Podsakoff, N. P. (2003). Common method biases in behavioral research: A critical review of the literature and recommended remedies. Journal of Applied Psychology, 88(5), 879–903.
• Raudenbush, S. W., & Bryk, A. S. (2002). Hierarchical linear models: Applications and data analysis methods (2nd ed.). SAGE.
• Rubin, D. B. (1987). Multiple imputation for nonresponse in surveys. Wiley.
• Schafer, J. L., & Graham, J. W. (2002). Missing data: Our view of the state of the art. Psychological Methods, 7(2), 147–177.
• Tong, A., Sainsbury, P., & Craig, J. (2007). Consolidated criteria for reporting qualitative research (COREQ): A 32-item checklist for interviews and focus groups. International Journal for Quality in Health Care, 19(6), 349–357.
• Willis, G. B. (2005). Cognitive interviewing: A tool for improving questionnaire design. SAGE.