Downloads: PDF · Markdown

Overview

How did Chile go from roughly 2% foreign-born in 2002 to over 8% in 2024? Where do immigrants settle, what work do they find, and do their health outcomes differ from those of Chilean-born residents?

The Census 2024 microdatos give us the most detailed snapshot ever of Chile’s population: over 19 million individual records linked to their households and dwellings. Two complementary health datasets enrich the picture:

• ENO (Enfermedades de Notificación Obligatoria), the national mandatory notifiable-disease surveillance registry, with about 333,000 records from 2007 to 2024.
• GRD (Grupos Relacionados de Diagnóstico), hospital discharge records covering about 5 million episodes from 2019 to 2024, coded with ICD-10 diagnoses.

In this project-driven course, students work in pairs to weave all three datasets into a coherent narrative: Migration, Health, and Socioeconomic Integration in Chile. Each team is assigned one to three comunas (mostly Región Metropolitana) chosen so that each team covers a similar total population. Each team builds a complete analytical pipeline: from raw parquet/CSV/pipe-delimited files through demographic profiling, disease surveillance analysis, hospital discharge analysis, spatial mapping, and finally cross-dataset ecological modeling at the comuna level.

Because the three datasets contain different people (a census respondent is not the same individual as an ENO notification or a hospital discharge), cross-dataset analysis is necessarily ecological: each dataset is aggregated to the comuna level and linked by codigo_comuna. Understanding when ecological inference is valid, and when it is not, is an explicit learning objective.

Inverted lectures

This is not a typical lecture course, by design. Each week students receive an assignment that introduces a specific skill or analytical technique. The structure is intentionally inverted: the instructor points to high-quality tutorials, documentation, and video content to study independently before class. Class time is spent diving deeper, debugging, and resolving roadblocks together. Engagement is
essential: come to class having reviewed the material, having tried the assignment, and with real questions ready.

Topics

The course covers, roughly in order:

• Tools. Jupyter (Colab / VS Code / JupyterLab), GitHub, Markdown; Python with pandas, pyarrow, geopandas, matplotlib, seaborn or plotly; statsmodels for regression.
• Census 2024 microdata. Vivienda / hogar / persona tables, linking keys (id_vivienda, id_hogar, id_persona), parquet, hierarchical joins, missing-value conventions (-99, NA).
• ENO. Semicolon-delimited CSV, 2007 to 2024; key fields: disease code, notification date, comuna, nationality, education; cleaning the “Desconocido” nationality category; rates over time.
• GRD. Pipe-delimited zipped files (one per year), 2019 to 2024; key fields DIAGNOSTICO1, COMUNA, nationality, length of stay, severity; ICD-10 lookup via CIE-10.xlsx with diagnostic chapter grouping.
• Demographic analysis. Age pyramids by sex, dependency ratios, household typology.
• Migration variables. p24_lug_resid5 (residence 5 years ago), p25_lug_nacimiento (place of birth), p26_llegada_periodo (arrival period), p27_nacionalidad.
• Labor force, education, housing. Labor force status, occupation (cod_ciuo), economic activity (cod_caenes), commute mode; years of schooling (escolaridad), education level (cine11); housing quality, overcrowding, internet access, tenure.
• Spatial analysis. Comuna-level mapping with geopandas + shapefiles; choropleths.
• Modeling. Logistic regression for binary census outcomes; Poisson and Negative Binomial for count outcomes (with population offset); ecological regression linking aggregate census predictors to health outcomes; coefficient interpretation (odds ratios, incidence rate ratios); predicted-rate maps; residual maps.

Prerequisites

• Taller de Programación en Python.
• IIP225A (Probability and Statistics).
• or instructor permission.

Assignments (summary)

All work is done in pairs, with each pair assigned one to three comunas of roughly similar total population. The three datasets are introduced progressively: Tarea 0 gives a shallow first contact with all three; Tarea 1 goes deep on Census; Tarea 2 goes deep on ENO + GRD; Tarea 3 merges everything at the comuna level.

• Tarea 0 (5 pts): Setup and first contact with all three datasets. Due Mar 12.
• Tarea 1 (10 pts): Demographic profile and migration landscape. Due Mar 26.
• Tarea 2 (10 pts): Health landscape (ENO + GRD). Due Apr 16.
• Quiz 1 (2 pts, bonus): Share comuna-level summary tables in a class-wide format. Due Apr 20.
• Tarea 3 (10 pts): Cross-dataset ecological modeling. Due Apr 30.
• Final project (25 pts): Pick one anomaly from your pipeline and defend it: GitHub repo + 8 to 10 minute video + handwritten in-person memo. See the Final project page.

Full assignment briefs (PDF and Markdown) are on the Assignments page.

Grading

You cannot pass the class without completing most of the core assignments and participating actively, as defined above.

Policies

• Late policy. 2-day extension available by email to staff before the deadline.
• Collaboration. Work is done in pairs. Both members must understand all code submitted.
• AI policy.
  • Tareas 0 to 3 and the final project’s async components (repo + video): AI tools are allowed. Disclose them in the README of your final repo.
  • Final project’s in-person memo: no AI, no laptops, no notes. This is the contract that gives the async work its meaning.
• Submission. Canvas (PDF of notebook + repo link for the final project; video link; the memo happens live).

Past instantiations

A similar course is offered often. See Fall 2025 and Spring 2025, or the parent course page.