Open Science
LCF 5900 · Reproducible and open processing of scientific data (ESALQ/USP, 2026)
The LCF 5900 course introduces open and accessible science: tools, software and databases that let other researchers access and reproduce your work. For the course I built three fully reproducible projects. Each one has open data, open code, a rendered report and a version you can run in the browser with Posit Cloud or Google Colab, with no installation needed.
Climate of Piracicaba and ENSO
How do long-term seasonal trends in temperature range and rainfall behave in Piracicaba since 1950, and how do El Niño / La Niña anomalies affect extreme local weather, especially summer rainfall? The workflow covers data extraction, exploratory analysis, Kruskal-Wallis tests, Spearman correlation, and forecasting of the Oceanic Niño Index (ONI) with ARIMA and Prophet.
GitHub Data Report (PDF) Run on Posit Cloud Colab notebook Course exercise
MAGIC common bean: reproducing a published analysis
Can a published genetics study be reproduced from its open data alone? I re-created the phenotypic and genotypic analyses of a MAGIC (Multiparent Advanced Generation Intercross) population of common bean (Phaseolus vulgaris L.) evaluated under drought, to show in practice that scientific reproducibility is feasible.
GitHub Data Report (PDF) Run on Posit Cloud Colab notebook Course exercise
Wheat breeding with Bayesian models and information theory
A reproducible pipeline for G×E in rainfed wheat trials. Bayesian hierarchical mixed models handle a 4×4 lattice design and missing data. Mutual information maps non-linear environmental networks and supports multi-trait selection, and Shannon entropy measures phenotypic stability as informational predictability. The result turns raw trial data into decision-support indicators without arbitrary index weights.
GitHub Data Report (PDF) Run on Posit Cloud Colab notebook Course exercise