Ankle Draft Risk Model Under Winter Clothing Conditions

Tobias Kramer, Junmeng Lyu, Stefano Schiavon
Center for the Built Environment, University of California, Berkeley

---

About

This repository contains the data processing code, statistical analysis, and manuscript source for the paper "Ankle draft risk model under winter clothing conditions", submitted to Building and Environment.

The study evaluates thermal comfort responses to ankle-level draft when participants wore winter clothing (0.75 clo). Based on controlled laboratory experiments with 51 participants, we find that the current ASHRAE Standard 55 ankle draft model overpredicts dissatisfaction by a factor of 2.8× under covered-ankle conditions, and propose separate draft risk models for covered and uncovered ankles.

---

Repository Structure

ankle-draft-II/
├── src/
│   ├── R/                  # Statistical analysis and figure generation
│   └── python/             # Raw data import and preprocessing
├── data/
│   ├── 00-raw/             # Raw sensor and survey files (not tracked by git)
│   ├── 01-processed/       # Cleaned, standardized data (Python output)
│   └── 02-export/          # Analysis outputs used in the manuscript
├── manuscript/
│   ├── index.qmd           # Main manuscript (Quarto)
│   ├── refs.bib            # Bibliography
│   ├── figs/               # Figures
│   └── tables/             # Results tables
└── docs/                   # Rendered HTML output

---

Analysis Pipeline

The analysis runs in two stages:

Stage 1 — Data preprocessing (Python)

Raw experimental data (environmental sensors, air velocity, skin temperature iButtons, and survey responses) are processed into clean CSV files.

Script	Purpose
src/python/import_metadata.py	Subject onboarding data
src/python/import_survey.py	Thermal comfort survey responses
src/python/import_enviromental.py	Environmental sensor logs (CO₂, temperature, humidity, etc.)
src/python/import_airflow.py	Air velocity measurements
src/python/import_tsk.py	Skin temperature (iButton) data
src/python/constants.py	Shared path and column-name configuration
src/python/survey_config.py	Survey question ID → variable name mapping

Stage 2 — Statistical analysis and figures (R)

R scripts are sourced in order by the manuscript at render time.

Script	Purpose
src/R/x_setup.R	Package loading, factor levels, color palettes, figure dimensions
src/R/x_data.R	Imports processed CSVs and builds tidy analysis dataframes
src/R/x_func.R	Reusable plotting and data-summarization functions
src/R/x_stat.R	Statistical helpers (odds ratios and CIs from mixed-effects models)
src/R/x_analysis.R	Main analysis: all figures, models, and results reported in the manuscript

Stage 3 — Manuscript rendering (Quarto)

manuscript/index.qmd sources the R analysis and renders to DOCX and HTML via Quarto.

quarto render manuscript/index.qmd

---

Reproducing the Analysis

Requirements

R (≥ 4.2) with renv for package management:

install.packages("renv")
renv::restore()

Python (≥ 3.10):

python -m venv .venv
source .venv/bin/activate
pip install -r requirements.txt

Quarto (≥ 1.4): quarto.org

Steps

1. Place raw data files in data/00-raw/ (see directory structure above).
2. Run the Python import scripts to populate data/01-processed/.
3. Open the RStudio project (ankle-draft-II.Rproj) or render from the terminal:

quarto render manuscript/index.qmd

The rendered manuscript is written to docs/.

---

Key Dependencies

Tool	Version
R	≥ 4.2
lme4	mixed-effects models
ggplot2 / patchwork	figures
gt	tables
Python	≥ 3.10
pandas	2.3.3
Quarto	≥ 1.4

Full R package versions are pinned in renv.lock. Python dependencies are in requirements.txt.

---

Citation

Kramer T, Lyu J, Schiavon S. Ankle draft risk model under winter clothing conditions. Building and Environment (under review).