Presentation Title:

From Field to Forecast: Building a Reproducible Analytical Pipeline for IOM DTM Data in R

This 45-minute presentation is structured to balance high-level humanitarian context with technical R implementation. It is designed to keep both “Domain Experts” and “R Developers” engaged by alternating between Why (the mission) and How (the code).

Total Duration: 45 Minutes

Format: 35 Min Presentation + 10 Min Q&A

I. Introduction: The Data-Driven Humanitarian (5 Minutes)

  • The Mission: Brief overview of IOM’s role in global migration.

  • The Problem: The “Data Silo” vs. “The Need for Speed.” Why manual spreadsheets fail in rapid-onset crises.

  • The Solution: An end-to-end R pipeline that ensures transparency, speed, and reproducibility.

II. Domain Architecture: What are we actually measuring? (7 Minutes)

  • The DTM Framework: explaining the “Four Pillars” (Mobility Tracking, Flow Monitoring, Registration, and Surveys).
  • Data Granularity: Understanding Admin Levels (0–2) and the concept of “Data Rounds.”
  • P-Codes: The “Primary Key” of the humanitarian world that links IOM data to the rest of the UN system.

III. Technical Architecture: Connecting R to the IOM API (8 Minutes)

  • The API Gateway: Introduction to DTM API v3.
  • The Authentication Layer: Using .Renviron to protect subscription keys.
  • The Ingestion Script: * Using httr2 for API requests.
  • Using jsonlite to flatten nested JSON structures into Tibbles.
  • Live Snippet: Showing a clean 10-line block of code that pulls live IDP figures.

IV. The Tidyverse Engine: Wrangling Humanitarian Data (10 Minutes)

  • Standardizing Chaos: Cleaning inconsistent place-names and handling NA values.
  • Longitudinal Logic: Using group_by() and lag() to calculate the velocity of displacement between rounds.
  • Demographic Reshaping: Using pivot_longer() to prepare Sex and Age Disaggregated Data (SADD) for population pyramids.
  • Vulnerability Indexing: Creating a weighted “Severity Score” using mutate() and case_when().

V. Spatial Analysis & Visualization (7 Minutes)

  • Mapping with sf: Joining DTM metrics with geographic shapefiles.
  • Effective Visuals: * Choropleth Maps: Visualizing IDP density.
  • Trend Lines: Tracking flow volumes over time.
  • The “So What?”: Moving from “The chart looks nice” to “This is where the mobile clinic needs to go.”

VI. Data Ethics: Security & Disclosure Control (3 Minutes)

  • The “Do No Harm” Principle: Why we never share raw microdata.
  • Statistical Disclosure Control (SDC): * A brief look at the sdcMicro package.
  • Techniques: K-anonymity and Local Suppression to protect vulnerable individuals.

VII. Conclusion & The Future of Analysis (5 Minutes)

  • Summary: How R bridges the gap between field enumerators and strategic decision-makers.
  • Looking Ahead: Toward Anticipatory Action—using R to link DTM data with climate or economic indicators for predictive modeling.

VIII. Q&A / Discussion (10 Minutes)

  • Open floor for technical questions (API limits, R packages) and domain questions (Data reliability, IOM methodology).

Presentation Notes for the Speaker:

  • For the R Experts: Remind them that “Admin 2” is essentially a JOIN key.

  • For the DTM Experts: Remind them that “Tidyverse” is essentially a more powerful, transparent version of an Excel Pivot Table.

  • Visual Aid: Keep code snippets on the left and the resulting “Humanitarian Map” on the right to show the direct impact of the script.