混合研究方法分析服务器

Quantitative Data Collection  →  QUAN Analysis
                                       ↓
                               INTEGRATION & COMPARISON
                                       ↓
Qualitative Data Collection   →  QUAL Analysis

Phase 1: QUANTITATIVE         Phase 2: QUALITATIVE
Survey Analysis        →      Follow-up Interviews
Identify Patterns      →      Explain Why Patterns Exist

         ↓
    INTEGRATION

Phase 1: QUALITATIVE          Phase 2: QUANTITATIVE
Interviews/Focus Groups →     Survey Development
Generate Themes         →     Test Themes at Scale

         ↓
    INTEGRATION

cd /home/tayyabcheema777/MMR_MCP/mixed-methods-mcp
./scripts/setup-env.sh
source .venv/bin/activate

python3 -m venv .venv
source .venv/bin/activate
pip install -e .
pip install -r requirements.txt

mixed-methods-mcp

/home/tayyabcheema777/MMR_MCP/mixed-methods-mcp/examples/
├── sample_survey.csv          ← Quantitative data
├── sample_interview_1.txt     ← Qualitative transcripts
├── sample_interview_2.txt
└── sample_interview_3.txt

# Create your research project structure
mkdir -p ~/my_dissertation_study/{data,interviews,outputs}

~/my_dissertation_study/
├── data/
│   ├── pilot_survey.xlsx           ← Your Excel files
│   ├── main_survey.csv             ← Your CSV files
│   └── demographics.xlsx
├── interviews/
│   ├── participant_001.txt         ← Interview transcripts
│   ├── participant_002.txt
│   ├── focus_group_1.txt
│   └── field_notes.txt
└── outputs/
    ├── reports/                    ← Generated reports
    ├── visualizations/             ← Charts and graphs
    └── codebooks/                  ← Coding schemes

# Through Claude Code, simply say:
"Load the quantitative data from ~/my_dissertation_study/data/main_survey.csv"
"Load all interview transcripts from ~/my_dissertation_study/interviews/"
"Export visualizations to ~/my_dissertation_study/outputs/visualizations/"

# For automatic quantitative processing
mkdir -p ~/auto_analysis/quant/{input,processed,reports}

# For automatic qualitative processing
mkdir -p ~/auto_analysis/qual/{input,processed,stage1,stage2,stage3}

# Step 1: Load and analyze your survey data (n=150 students)
"Load the survey data from ~/dissertation/student_survey.xlsx"
# System responds with variable summary and next steps

# Step 2: Explore relationships
"Show me descriptive statistics for self_efficacy and engagement scores"
"Create a correlation heatmap for all variables"
"Test if high vs low self-efficacy groups differ in engagement"

# Step 3: Select interview participants (Explanatory Sequential)
"Use purposeful sampling to select 6 students with extreme engagement scores"
# System identifies 3 high and 3 low engagement students

# Step 4: Load and code interview transcripts
"Load all interview files from ~/dissertation/interviews/"
"Search for text segments mentioning motivation or barriers"
# Review results, then code them

"Apply code 'intrinsic motivation' to document participant_high_1,
 characters 450-680, with memo 'student describes genuine interest in subject'"

# Step 5: Create integrated findings
"Merge the survey data with interview codes using participant_id"
"Create a joint display showing engagement scores alongside motivation themes"
"Compare patterns between quantitative engagement and qualitative themes"

# Step 6: Generate outputs
"Export correlation heatmap to ~/dissertation/outputs/figures/"
"Export integrated report to ~/dissertation/outputs/chapter4_findings.docx"
"Save my current research context"

# Step 7: Document your process
"Log decision: 'Used median split for high/low engagement groups'
 with rationale 'balanced sample sizes needed for qualitative follow-up'"

# Convergent Design - analyze both data sources in parallel

# Quantitative strand
"Load pre and post program survey data from ~/program_eval/surveys.xlsx"
"Compare pre_test vs post_test scores for teacher confidence"
"Identify outliers in the improvement scores"
"Create comparison chart for pre vs post across all domains"

# Qualitative strand (at the same time)
"Load focus group transcripts from ~/program_eval/focus_groups/"
"Apply codes for themes like 'practical_strategies', 'peer_collaboration',
 'time_constraints', 'curriculum_alignment'"
"Generate word cloud to see most frequent terms"
"Show theme frequency to identify most common themes"

# Integration
"Create joint display with teacher_id, improvement_score, and primary themes"
"Compare patterns - do teachers with high improvement mention different themes?"
"Export integrated report documenting both strands' findings"

# Phase 1: Qualitative exploration
"Load initial interview transcripts from ~/tech_adoption/phase1_interviews/"
"Apply open codes as themes emerge from the data"
"Check code co-occurrence to find related themes"
"Generate theme network to visualize relationships"
"Quantitize codes to see frequency distribution"

# Phase 2: Survey development (based on Phase 1 themes)
"Load Phase 2 survey data testing identified themes at scale"
"Test correlations between the factors identified in Phase 1"
"Use regression to identify strongest predictors"

# Integration
"Compare patterns between qualitative themes and quantitative factors"
"Create integration memo documenting how qual themes were confirmed/refined"

┌─────────────────────────────────────────────────────────────┐
│                    YOUR RESEARCH WORKFLOW                    │
└─────────────────────────────────────────────────────────────┘

   QUANTITATIVE STRAND              QUALITATIVE STRAND
          ↓                                ↓
┌──────────────────────┐      ┌──────────────────────┐
│ Load Survey Data     │      │ Load Interview Files │
│ (.xlsx, .csv)        │      │ (.txt, .docx)        │
└──────────┬───────────┘      └──────────┬───────────┘
           │                              │
           ↓                              ↓
┌──────────────────────┐      ┌──────────────────────┐
│ Descriptive Stats    │      │ Read & Code Text     │
│ • Mean, SD           │      │ • Apply themes       │
│ • Distributions      │      │ • Write memos        │
└──────────┬───────────┘      └──────────┬───────────┘
           │                              │
           ↓                              ↓
┌──────────────────────┐      ┌──────────────────────┐
│ Inferential Tests    │      │ Thematic Analysis    │
│ • Correlations       │      │ • Theme frequency    │
│ • Group comparisons  │      │ • Co-occurrence      │
│ • Regression         │      │ • Word clouds        │
└──────────┬───────────┘      └──────────┬───────────┘
           │                              │
           └──────────────┬───────────────┘
                          ↓
            ┌──────────────────────────┐
            │   INTEGRATION PHASE      │
            │  • Merge by ID           │
            │  • Joint displays        │
            │  • Pattern comparison    │
            │  • Integration memos     │
            └─────────────┬────────────┘
                          ↓
            ┌──────────────────────────┐
            │    FINAL OUTPUTS         │
            │  • Integrated report     │
            │  • Publication figures   │
            │  • Codebook              │
            │  • Research context      │
            └──────────────────────────┘

# Set up your folders
mkdir -p ~/auto_coding/{input,processed,outputs}

# Start the automated watcher
mixed-methods-mcp-cli watch ~/auto_coding/input \
  --processed-dir ~/auto_coding/processed \
  --stage1-cmd "your-ai-command-stage1" \
  --stage2-cmd "your-ai-command-stage2" \
  --stage3-cmd "your-ai-command-stage3"

# Drop your transcript files (.csv or .xlsx) into ~/auto_coding/input/
# The system processes them automatically every 30 seconds

# Set up folders
mkdir -p ~/auto_stats/{input,processed,reports}

# Start the watcher
mixed-methods-mcp-cli quant-watch ~/auto_stats/input \
  --processed-dir ~/auto_stats/processed \
  --reports-dir ~/auto_stats/reports \
  --llm-cmd "your-statistical-advisor-command"

# Place your dataset in ~/auto_stats/input/
# Review recommended tests
# Approve or decline
# Receive comprehensive statistical report

"What's my current research context status?"

{
  "quantitative_data": "~/dissertation/survey_n150.xlsx",
  "loaded_variables": ["participant_id", "self_efficacy", "engagement", "gender", "year"],
  "qualitative_documents": 6,
  "document_names": ["participant_001.txt", "participant_002.txt", ...],
  "coded_segments": 45,
  "unique_codes": 8,
  "codes_used": ["intrinsic_motivation", "time_management", "peer_support", ...],
  "decisions_logged": 3,
  "suggestions": [
    "Consider running correlation analysis on self_efficacy and engagement",
    "You have both data types - ready to create joint display",
    "Export your codebook before proceeding to integration"
  ]
}

# Start with the samples to learn the system
"Load examples/sample_survey.csv for practice"
"Show me correlations between competence, confidence, and engagement"
"Create a correlation heatmap"

"Load all sample interviews from examples/ folder"
"Search for mentions of 'challenge' or 'difficult'"
"Apply code 'learning_barriers' to the relevant segments"
"Generate word cloud from all interviews"
"Show theme frequency"

cd /home/tayyabcheema777/MMR_MCP/mixed-methods-mcp
source .venv/bin/activate

# Instead of: "examples/data.csv"
# Use: "/home/tayyabcheema777/MMR_MCP/mixed-methods-mcp/examples/data.csv"
# Or: "~/MMR_MCP/mixed-methods-mcp/examples/data.csv"

rm -rf ~/.mixed_methods_mcp
# The system will recreate it automatically

pip install kaleido
# Or re-run setup: ./scripts/setup-env.sh

rm ~/.mixed_methods_mcp/project_context.json
# The system creates a fresh one

"Log decision: Used median split for creating high/low groups
 with rationale: Balanced sample sizes needed for qualitative follow-up
 with alternatives: Could have used tertile split or quartiles"

# After major analysis steps:
"Save my research context"
# This protects your work and creates timestamps

"Export codebook to ~/my_study/codebook_v1.docx"
# Update version numbers as your coding evolves

"Create integration memo with these highlights:
 - High anxiety students mentioned time management issues (convergence)
 - Low anxiety students described peer support (qualitative expansion)
 - Gender differences in coping strategies (divergence - needs explanation)"

# Check your history before major decisions
"Show me my analysis history for the last 20 steps"

# Create a workflow YAML file: my_workflow.yaml
# Then execute it:
mixed-methods-mcp-cli run my_workflow.yaml --save-output results.json

steps:
  - tool: load_quantitative_data
    args:
      path: data/survey.xlsx

  - tool: describe_statistics

  - tool: correlation_analysis
    args:
      method: pearson

  - tool: load_documents
    args:
      path: interviews/
      pattern: "*.txt"

  - tool: create_joint_display
    args:
      id_column: participant_id
      metrics: ["anxiety_score", "engagement_score"]

from mixed_methods_mcp import QuantitativeToolkit, QualitativeToolkit

# Use toolkits directly in your Python scripts
quant = QuantitativeToolkit()
quant.load_quantitative_data("data.xlsx")
stats = quant.describe_statistics()

qual = QualitativeToolkit()
qual.load_documents("interviews/")
qual.apply_code("interview1.txt", "coping_strategy", 100, 250)

Ilyas, M. T. (2025). Mixed Methods MCP Server (Version 0.1.0) [Computer software].
https://github.com/your-repo/mixed-methods-mcp

All data analysis was conducted using the Mixed Methods MCP Server
(Ilyas, 2025), an AI-assisted research analysis platform implementing
procedures described by Creswell and Plano Clark (2018).