Replication Data for: \"Promoting Sustainable Travel Modes Through Health and Active Lifestyle Messaging\"

Data and Code Repository This repository contains the anonymized dataset and analysis code associated with the paper titled \"Promoting Sustainable Travel Modes Through Health and Active Lifestyle Messaging.\" Contents anonymized_data.csv: Contains the raw anonymized dataset (1 MB) used in the analysis. analysis_v1.ipynb: Python scripts for analysis, and visualization. (Last Modified July 7, 2025, 5:00PM) README.md: Description of the repository contents and usage. datadictionary.md: A detailed explanation of each variable in the final dataset. 68 unique variables, 4,840 observations. Requirements Packages required to run this analysis are pandas==2.0.3, numpy==1.24.1, statsmodel.api==0.14.1. This code was tested on Python 3.8.13 and 3.9.2 and on macOS Sequoia 15.5 and Google Colab CPUs. Structure of the code The first code block loads the dataset and required packages. The second code block has helper function that generates dataframe for statistical analysis in the later blocks. The third code block has helper variables and functions to load model specifications and formatting model coefficients for analysis in the later blocks Code blocks four and above generate statistical results used in the paper. Output This code package generates the necessary derived data consisting of odds ratios and uncertainty for Figure 1 and Figure 2 in the main document: Main Document Fig. 1 Treatment effects of air quality impacts targeting bus transit and active lifestyle messaging targeting walking or biking. Main Document Fig. 2 Comparative treatment effects of health and active lifestyle messaging for all respondents and those with underlying health conditions. This code package generates the following tables: Main Document Table 1 Heterogeneous treatment effects across key subgroups. Table S4. Treatment effects of air pollution exposure messaging in Experiment 1 (personal and community benefits targeting bus transit) Table S5. Treatment effects of air quality improvement messaging in Experiment 2 (personal and community benefits targeting bus transit) Table S6. Treatment effects of air quality improvement messaging in Experiment 2 (personal gain targeting walking) Table S7. Treatment effects of air quality improvement messaging in Experiment 2 (personal and community benefits targeting walking) Table S8. Treatment effects of air quality improvement messaging in Experiment 2 (personal gain targeting biking) Table S9. Treatment effects of air quality improvement messaging in Experiment 2 (personal and community benefits targeting biking) Table S10. Treatment effects of all active lifestyle messaging in Experiment 3 (personal and community benefits targeting walking and biking) Table S11. Treatment effects of all active lifestyle messaging in Experiment 3 (personal gain targeting biking) Table S12. Treatment effects of active lifestyle messaging in Experiment 3 (personal and community benefits targeting biking) Table S13. Treatment effects of step count messaging in Experiment 3 (personal and community benefits targeting walking) Table S14. Treatment effects of calories burned messaging in Experiment 3 (personal and community benefits targeting walking) Table S15. Treatment effects of heart health messaging in Experiment 3 (personal and community benefits targeting walking and biking) Table S16. Heterogeneous treatment effects of air pollution exposure messaging in Experiment 1 (personal gain targeting bus transit) Table S17. Heterogeneous treatment effects of air pollution exposure messaging in Experiment 1 (personal and community benefits targeting bus transit) Table S18. Heterogeneous treatment effects of air quality improvement messaging in Experiment 2 (personal gain targeting bus transit) Table S19. Heterogeneous treatment effects of air quality improvement messaging in Experiment 2 (personal and community benefits targeting bus transit) Table S20. Heterogeneous treatment effects of active lifestyle messaging in Experiment 3 (personal gain targeting walking) Table S21. Heterogeneous treatment effects of active lifestyle messaging in Experiment 3 (personal and community benefits targeting walking) Table S22. Heterogeneous treatment effects of different active lifestyle messaging in Experiment 3 targeting walking Table S23. Heterogeneous treatment effects of calories burned messaging for commuters with varying daily travel times in Experiment 3 (targeting walking) Replication Supporting replication code is also available here: https://github.com/asensio-lab/health-active-lifestyle. This code package was last replicated on July 7, 2025 by @YifanLiu0304 Declaration of generative AI and AI-assisted technologies in the coding process During the preparation of this work the authors used ChatGPT in order to debug code errors such as KeyError, syntax errors in python scripts that were used to generate statistical tables. After using this tool/service, the researchers reviewed, edited, and replicated all code.

数据与代码仓库
本仓库包含与题为《通过健康与积极生活方式宣传推广可持续出行方式》的论文相关的匿名化数据集与分析代码。

### 仓库内容
- `anonymized_data.csv`：包含本研究分析所用的匿名化原始数据集（1 MB）。
- `analysis_v1.ipynb`：用于分析与可视化的Python脚本（最后更新时间：2025年7月7日，17:00）。
- `README.md`：本仓库的内容说明与使用指南。
- `datadictionary.md`：对最终数据集中各变量的详细说明，共包含68个唯一变量与4840条观测样本。

### 依赖要求
本分析所需的Python依赖包为：`pandas==2.0.3`、`numpy==1.24.1`、`statsmodel.api==0.14.1`。本代码已在Python 3.8.13、3.9.2版本，以及macOS Sequoia 15.5系统与Google Colab CPU环境下完成测试。

### 代码结构
1. 首个代码块：加载数据集与所需依赖包；
2. 第二个代码块：包含辅助函数，用于生成后续统计分析所需的数据集；
3. 第三个代码块：包含辅助变量与函数，用于加载模型设定并格式化模型系数，以供后续分析使用；
4. 第四个及后续代码块：用于生成论文所需的各类统计结果。

### 输出结果
本代码包可生成论文主文档中图1与图2所需的派生数据，包含比值比（odds ratios）与不确定性区间：
- 主文档图1：针对公共交通的空气质量影响宣传，以及针对步行或骑行的积极生活方式宣传（active lifestyle messaging）的处理效应（treatment effects）。
- 主文档图2：针对全体受访者与伴有基础性健康状况的人群的健康及积极生活方式宣传的对比处理效应。

本代码包可生成如下表格：
1. 主文档表1：关键亚组间的异质性处理效应（heterogeneous treatment effects）。
2. 表S4：实验1中空气污染暴露宣传的处理效应（针对公共交通的个人与社区收益导向宣传）
3. 表S5：实验2中空气质量改善宣传的处理效应（针对公共交通的个人与社区收益导向宣传）
4. 表S6：实验2中空气质量改善宣传的处理效应（针对步行的个人收益导向宣传）
5. 表S7：实验2中空气质量改善宣传的处理效应（针对步行的个人与社区收益导向宣传）
6. 表S8：实验2中空气质量改善宣传的处理效应（针对骑行的个人收益导向宣传）
7. 表S9：实验2中空气质量改善宣传的处理效应（针对骑行的个人与社区收益导向宣传）
8. 表S10：实验3中所有积极生活方式宣传的处理效应（针对步行与骑行的个人与社区收益导向宣传）
9. 表S11：实验3中所有积极生活方式宣传的处理效应（针对骑行的个人收益导向宣传）
10. 表S12：实验3中积极生活方式宣传的处理效应（针对骑行的个人与社区收益导向宣传）
11. 表S13：实验3中步数宣传的处理效应（针对步行的个人与社区收益导向宣传）
12. 表S14：实验3中热量消耗宣传的处理效应（针对步行的个人与社区收益导向宣传）
13. 表S15：实验3中心脏健康宣传的处理效应（针对步行与骑行的个人与社区收益导向宣传）
14. 表S16：实验1中空气污染暴露宣传的异质性处理效应（针对公共交通的个人收益导向宣传）
15. 表S17：实验1中空气污染暴露宣传的异质性处理效应（针对公共交通的个人与社区收益导向宣传）
16. 表S18：实验2中空气质量改善宣传的异质性处理效应（针对公共交通的个人收益导向宣传）
17. 表S19：实验2中空气质量改善宣传的异质性处理效应（针对公共交通的个人与社区收益导向宣传）
18. 表S20：实验3中积极生活方式宣传的异质性处理效应（针对步行的个人收益导向宣传）
19. 表S21：实验3中积极生活方式宣传的异质性处理效应（针对步行的个人与社区收益导向宣传）
20. 表S22：实验3中针对步行的不同积极生活方式宣传的异质性处理效应
21. 表S23：实验3中针对步行的热量消耗宣传对不同每日出行时长通勤者的异质性处理效应

### 可复现性支持
本研究的复现代码可通过以下链接获取：https://github.com/asensio-lab/health-active-lifestyle。本代码包于2025年7月7日由@YifanLiu0304完成复现测试。

### 编码过程中生成式AI与AI辅助技术声明
在本研究筹备期间，作者使用ChatGPT调试代码错误，例如用于生成统计表格的Python脚本中的KeyError与语法错误。使用该工具/服务后，研究人员已对所有代码进行审核、编辑与复现。