Education Sampling Data Management

Education Sampling Data Management

Goal

Create a practical sampling, recruitment, participant-ID, consent, anonymization, and data-management plan for education research.

Use After

Use after a study path and basic research design are selected:

• education-quantitative-study-design
• education-qualitative-study-design
• education-mixed-methods-design
• education-action-research-design
• education-design-based-research
• education-program-evaluation
• education-psychometric-scale-development
• education-learning-analytics-design

Do not expose the skill name to users. Present it as "样本与数据管理设计".

Inputs

• Paper type and research design
• Target participants: students, teachers, parents, administrators, schools/classes
• Education level, subject, grade, region
• Data sources: surveys, tests, interviews, observations, student work, logs, documents
• Whether minors are involved
• Access constraints: school permission, class access, teacher cooperation
• Data sensitivity and sharing requirements

Workflow

1. Identify sampling unit:
   • student
   • teacher
   • class
   • school
   • lesson/session
   • artifact/document
   • platform account/log
2. Select sampling strategy:
   • convenience
   • purposive
   • stratified
   • cluster/class-based
   • random
   • snowball
   • maximum variation
   • theoretical sampling for grounded theory
3. Define inclusion/exclusion criteria.
4. Estimate sample size needs:
   • quantitative: power/precision/model complexity
   • qualitative: information richness/saturation
   • psychometric: item-to-participant and factor analysis needs
   • meta/log data: unit count and time window
5. Design recruitment workflow:
   • school/administrator permission
   • teacher coordination
   • parent/guardian consent if minors
   • student assent when appropriate
   • participant reminders
6. Create anonymized participant ID scheme.
7. Create data dictionary and linking table plan.
8. Define secure storage, file naming, versioning, and access control.
9. Plan missing-data and attrition tracking.
10. Output a data-management package.

Tool Calls

Sample Size and Power

G*Power: https://www.psychologie.hhu.de/arbeitsgruppen/allgemeine-psychologie-und-arbeitspsychologie/gpower

R packages:

install.packages(c("pwr", "simr", "WebPower"))

Participant/Data Tracking

Excel / Google Sheets
Airtable: https://www.airtable.com/
REDCap: https://www.project-redcap.org/
Notion: https://www.notion.so/
Open Science Framework: https://osf.io/

Secure Survey/Data Collection

Qualtrics: https://www.qualtrics.com/
问卷星: https://www.wjx.cn/
Google Forms: https://forms.google.com/
REDCap: https://www.project-redcap.org/

Output Format

1. Sampling Plan

2. Recruitment Workflow

3. Participant ID Scheme

Use anonymous IDs, not names.

Example:

4. Data Dictionary

Example levels:

• student-level
• teacher-level
• class-level
• school-level
• session-level
• item-level
• artifact-level

5. Dataset Inventory

6. Consent and Ethics Checklist

Recommended File Naming

projectname_dataset_unit_date_version.ext
aiwriting_survey_student_20260609_v01.xlsx
aiwriting_interview_teacher_20260609_v01.docx
aiwriting_score_student_prepost_20260609_v01.xlsx

Data-Linking Rule

Maintain two layers:

1. Analysis datasets with anonymous IDs only.
2. Separate identity linking table stored securely and not included in analysis exports.

Never put real names, phone numbers, parent contacts, student numbers, or school-identifying details in the analysis dataset unless there is a clear approved reason.

Education-Specific Risks

• Minors require extra consent/assent attention.
• Class-based sampling often creates nested data; record class/school IDs.
• Intervention studies need attrition tracking by group.
• Teacher-collected data may create pressure on students; make voluntariness explicit.
• AI platform logs may include sensitive prompts, writing samples, or account identifiers; anonymize text artifacts carefully.

Quality Rules

• Every dataset must have a unique participant or unit ID.
• Record group assignment, time point, and data source early; do not reconstruct later from memory.
• Keep raw data, cleaned data, and analysis data separate.
• Use consistent missing-value codes.
• Record all data exclusions with reasons.
• For multi-level education data, preserve school/class/teacher/student hierarchy.

User-Facing Closure

End by asking for the most important missing feasibility detail:

接下来需要确认一个实际条件：你预计能接触几个班、多少名学生/教师？我可以据此帮你把样本表、匿名编号规则和数据字典做成可直接采集的版本。