How lucylgao/clusterpval-experiments works
• Simulation Code: Represented by the directory “simulation-code”, this component contains various R scripts (e.g., naive_type1.R, our_type1.R, cond_power.R, etc.) that run simulation experiments.
• Simulation Results: Represented by “simulation-results”, these Rdata files are generated by running the scripts in simulation-code.
• Figures Generation: Two parts here:
• “figures-code” includes the R scripts that actually call the code to produce the final figures.
• “figures” holds the output PDFs that are used in the paper.
• Real Data Analysis: Represented by “real-data-code”, where scripts such as penguins.R and zheng.R perform analyses on real external datasets (with external data downloaded from provided links).
Data and Workflow Relationships
• Simulation Workflow:
• The simulation scripts in “simulation-code” run to produce simulation data outputs.
• These outputs are stored as “simulation-results” which are then consumed by either the figures-code or directly referenced when producing figures.
• Figures Generation Workflow:
• The “figures-code” uses the simulation results (and in some cases additional processing) to produce publication-grade figures that are stored in the “figures” directory.
• Real Data Analysis Workflow:
• The “real-data-code” scripts import external datasets, process them, and generate respective analysis results and potentially figures (although the final figures for the paper seem to be separated).
Architectural Patterns & Design Principles
• Pipeline-driven Execution: The project is organized as a sequence of processing steps – simulation generation, results storage, and figures production – similar to a scientific data pipeline.
• Modularity: Each directory (simulation, real-data, figures) clearly represents a separate concern. The simulation-code and real-data-code are self-contained and independent yet feed into a central figure generation process.
• Reproducibility: The design emphasizes a reproducible workflow, where outputs (results and figures) are directly traceable back to the source R scripts.
Relevant Technologies
• R Language: All components are realized through R scripts and R-specific data files (.Rdata, .Rhistory) which hint at statistical computing frameworks.
• External Data Sources: In real-data-code, external links for datasets (penguins, T-cell, B-cell, monocyte data) are critical dependencies.
• File-based Workflow: The use of distinct directories for code, results, and figures naturally lends itself to a file-based data pipeline architecture.