thepolicylab/covid-vaccinessocialnorms

Code associated with social norms paper

Social Norms/Perceptions and Vaccine Intentions

This project uses a series of surveys run by YouGov to learn about the relationship between perceptions of the vaccination intentions of others and one's own vaccination intentions.

Requirements

Software

The analyses for this project were created in R version 4.1.0. You will also need the GLPK system library, which you can install with

sudo apt-get install libglpk-dev  # On Debian/Ubuntu
brew install glpk                 # On MacOS

To install the R packages used here, run the following command after cloning the repository.

Rscript -e 'if(!requireNamespace("remotes")){install.packages("remotes");remotes::install_github("rstudio/renv")} else {remotes::install_github("rstudio/renv")}'

Once these are installed, you can install most dependencies with

Rscript -e 'renv::restore()'
webshot::install_phantomjs()

A Note on Gurobi

This project involves the creation of a paired research design using non-bipartite matching using the nmatch command in the designmatch package for R. For speed and accuracy, we follow the advice of that package to use its interface to the Gurobi solver. We are using the free academic license. If you have trouble installing Gurobi for R, there are a few paths forward:

• the authors of designmatch have some older instructions that may help debugging here
• if you set the environment variable VACCINE_OPTIMIZER=GLPK the code will attempt to use the GPLK package instead. This will likely yield slightly different answers than those found in our work.

Due to the proprietary nature of this software, we've not been able to include it in our lockfile. The version we used in this analysis is 9.51 on MacOS 11. It produced identical outcomes on both M1 and Intel based Macs. However, earlier verisons of Gurobi do produce slightly different results. If you are attempting to replicate this project, please take note of the Gurobi version.

Feel free to raise a Github Issue or email our team if you'd like help with this or any other aspect of reproducing our findings.

Running the primary analysis

Once all of this is setup, then the main analysis can be run using the following command:

make all

Exploratory analysis: SPSS

Several exploratory analyses were performed using SPSS version 25. To run them, see instructions in the folder src/SPSS.

Exploratory analysis: Duplicates in Python

While revising this study, we discovered several people responded to both of our survey instruments. Ultimately, we decided to keep the first in time responses for reasons laid out in the paper. However, if you are interested in demographic characterisitics of duplicates across time, see src/notebooks/exploring_duplicates.ipynb.

To run this analysis, we utilized [poetry] and Python 3.8.6. To install poetry, see their instructions. Once installed, you can run

poetry install               # Install dependencies
poetry run jupyter notebook  # Start jupyter to run the notebook

Code layout

Our code can generally be divided into two sections:

• Data preparation
• Analysis
• Supplemental code

Data Preparation

Data preparation occurred using SPSS. Final datasets in .csv format were placed into data and are used in the analysis.

Some further processing occurred in R to handle duplicates. This occurs in 005_make_id_crosswalk.R in which we subset the main data set (MERGE_NR_2.3.21.csv) into first in time responses to our surveys (deduplicated_data_first.csv) and last in time responses to our surveys (deduplicated_data_last.csv).

Analysis

The analysis occurred using R. The files to produce the analyses are in src/R. In general, the analysis is broken down as follows:

• 00x: Configuration and a bit of data preparation
• 01x: Study 1 matching code
• 02x: Study 2 matching code
• 03x: Outcome analysis
• 06x: Figure generation
• 99x: No longer applicable code that may be of interest
• *.Rmd: Final analysis documents. Data for the paper appears here

To make the report and all associated figures (each in their own file src/R/figures) do make allfigs.

See the Makefile for the dependencies between files. Also see the included figure. Note that you can remake this figure by running make workflowplots though this will require Python 3.7.1+ and graphviz to be installed.

Supplemental code

In src/tplutils you'll find utilities for making the workflow plot.

In src/notebooks you'll find some simple summary statistics of folks who took both of our relevant surveys.

License

MIT