Post-hoc/Secondary Analysis Study Lifecycle

This guide walks through the complete process of conducting a post-hoc or secondary research with SafeInsights.

[Diagram showing 8 steps with estimated timelines]

Step 1: Define Your Research Question (1-2 weeks)

Start with a clear, focused research question that can be answered with observational data.

Good examples:

  • "How does frequency of practice predict final exam scores in college algebra?"
  • "What patterns of resource use are associated with course completion?"

Not yet supported:

  • "Does my new intervention improve learning?" (requires A/B test)
  • "How do OpenStax and Khan Academy compare?" (requires fusion)

Resources:


Step 2: Check Data Feasibility (1-2 weeks)

Verify that the data you need exists and is accessible.

Actions:

  1. Review the Data Catalog for relevant members
  2. Check what data types are available (events, assessments, demographics)
  3. Review any limitations or constraints
  4. Access simulated data to understand data structure

Key Questions:

  • Does a member have the variables I need?
  • Is the sample size adequate?
  • Are there temporal coverage gaps?
  • What demographic data is available?

Resources:

  • Organization-specific getting started guides in Data Catalog
  • Simulated data (links in each org's section)

Step 3: Draft Your Study Concept (1-2 weeks)

Develop a detailed study design.

What to include:

  • Research question and hypotheses
  • Target population and sample
  • Variables and measures
  • Analytic approach
  • Expected outputs

Resources:


Step 4: Submit Proposal (Day 1)

Submit your proposal through the SafeInsights Management App.

[Link to Management App]

Your proposal will be reviewed by:

  1. SafeInsights governance (privacy, FEAT compliance)
  2. Member organization (data access policies, priorities)

Step 5: Proposal Review (2-4 weeks)

Both SafeInsights and the member organization review your proposal.

Review criteria:

  • Scientific merit and feasibility
  • Privacy protections adequate
  • FEAT compliance
  • Alignment with member priorities
  • Outputs meet privacy thresholds

Possible outcomes:

  • ✅ Approved → Proceed to Step 6
  • 🔄 Revisions needed → Address feedback and resubmit
  • ❌ Declined → Receive explanation and guidance

Step 6: Develop & Test Container (2-6 weeks)

Write your analysis code and package it as a container.

Actions:

  1. Download simulated data for your target member
  2. Write analysis code (Python, R, or SQL)
  3. Test locally with simulated data
  4. Package as container
  5. Document your code
  6. Submit for review

Resources:

Code Review: Member organization reviews your code for:

  • Security (no data exfiltration attempts)
  • Privacy compliance
  • Correct use of data

Step 7: Container Execution (Minutes to hours)

Once approved, your container runs on real data within the secure enclave.

What happens:

  • Container spins up in isolated environment
  • Reads data (read-only access)
  • Executes your analysis
  • Generates outputs
  • Outputs queued for review

Duration: Depends on data volume and complexity (typically minutes to hours)


Step 8: Output Review & Release (1-2 weeks)

Before you receive results, outputs are reviewed to ensure privacy compliance.

Review process:

  • Automated checks for minimum cell sizes
  • Human review for any edge cases
  • Verification that outputs match approved proposal

You receive:

  • Aggregated statistics
  • Summary tables
  • Statistical model results
  • Plots and visualizations

You do NOT receive:

  • Individual-level data
  • Small cell counts
  • Any identifiable information

Resources:


Step 9: Analyze & Publish (Ongoing)

Use your approved outputs for analysis and publication.

Publishing considerations:

  • Cite SafeInsights appropriately
  • Acknowledge the data organization
  • Share code and analysis approach (where possible)
  • Consider preregistration (optional)

Typical Timeline

PhaseDuration
Feasibility & Design2-4 weeks
Proposal Review2-4 weeks
Container Development2-6 weeks
Container ExecutionHours to 1 day
Output Review1-2 weeks
Total~2-4 months

Tips for Success

  1. Start with feasibility: Don't invest time in a full proposal before checking data availability
  2. Use simulated data early: Develop code locally before submission
  3. Be specific in proposals: Vague proposals take longer to review
  4. Plan for iteration: First studies often need revision—that's normal
  5. Engage early: Use support resources to get feedback

Need Help?