Data
All scripts to reproduce our analyses can be found under
preregistrations
, grouped per study. Each study sub-folder
contains an analysis objects
folder, which contains both
the raw and cleaned data files. It is therefore possible to immediately
run the primary analyses scripts (e.g.,
1_primary_analyses.R
) without preprocessing the data first.
However, if you’d like to go through the analysis pipeline
start-to-finish, you should take the following steps:
1. Loading Dependencies
dependencies.R
:
Loads (and installs if necessary) all the necessary R packages as they
were on 2024-01-01 using the Groundhog
package. This
ensures that scripts will continue functioning as intended even after
packages have been updated. Groundhog
is a more
lightweight, easy-to-use alternative to package dependency managers such
as renv
.
2. Data Preparation
Each study script
folder contains a subfolder names
0_data_prep
. This subfolder contains all the scripts
necessary to get from the (minimally) preprocessed data to the cleaned
data. The scripts are numbered, and should be run in that order.
Importantly, the 1_create_scales.R
script will not work, as
it requires access to the raw data. We do not share these raw data files
as they contain variables that are not necessary for the analyses, as
well as identifying information (e.g., Prolific IDs). So, data
preprocessing should start at 2_clean_self_report.R
.
A note on running DDM models. Please note that
rerunning the DDM models takes a substantial amount of time, in
particular the Hierarchical Bayesian DDM models and the SSP models. The
SSP model in particular, takes around 20-30 minutes per participant to
complete. Even in parallel, this will take multiple hours. Thus, in most
cases, it might be better to use the DDM result objects in each study’s
analysis_objects
folder as a starting point.
Direct links to:
- Pilot
preprocessing scripts.
- Study
1 preprocessing scripts.
- Study
2 preprocessing scripts.
3. Primary Analyses
The script for the primary analyses for each study can be found under
preregistrations/<study>/scripts
. This script reads
in the cleaned data, runs all the analyses, and stores the main results
as an .RData file under
preregistrations/<study>/analysis_objects
.
4. Manuscript
manuscript/scripts/staging.R
:
This folder contains four scripts: one for each separate study, and one
for the pooled analyses. These scripts load the .RData files stored in
the analysis_objects
folder, and states the statistics that
are reported in the manuscript (e.g., creating Figures and Tables, and
formatting descriptive statistics). The results per study are stored in
a single .RData file under manuscript
(e.g., study1_staged_results.RData
).
manuscript/manuscript.qmd
:
loads the staged results and knits together the manuscript and all
statistics, tables, and figures computed in previous scripts.
5. Supplement
supplement/supplements.qmd
:
loads the staged results list and knits together the written supplement
and all tables and figures computed by the staging scripts.
6. Issues with dependencies?
Even if you use the dependencies.R
file to reinstate the
correct versions of packages, the code might break over time as a result
of changing operation systems or incompatibilities with future R
versions. To solve, this, we supply a Docker image that instantiates an
Rstudio cloud environment with all the required dependencies.
In short, Docker provides a ‘container’—similar to a virtual
machine—with all the project dependencies (R packages, software,
operating systems) pre-loaded in their (historic) versions as they were
used when the project was created. Thus, Docker ensures that scripts
will still function as intended even into the future. Note: It
is not necessary to use Docker to reproduce our results. If
you’re cloning the repository within a couple of years of publication
and installing the proper dependencies (see step 1 below) you should be
able to run all scripts without issues. However, some of our scripts
might stop to function properly over time as other dependencies get
updated and/or deprecated (e.g., future versions of RStudio) If that
case, Docker is the solution.
Below is the exact order these scripts should be ran (irrespective of
whether you are working with or without Docker):
If you want to create a Docker container in which to run the scripts,
you will have to follow the following steps:
- Download and install docker at https://docs.docker.com/get-docker/.
- If on Windows, open the PowerShell. If on Mac, open the terminal
through ‘Applications > Utilities > Terminal’.
- On the command line, type:
docker run --rm -d -e PASSWORD=my_password -p 8787:8787 stefanvermeent/attention_project
- Open a browser window and enter:
localhost:8787
as the
URL.
- You will be redirected to an Rstudio cloud login page. As the
username, type rstudio. As the password, type
my_password, unless you changed it under step 3.
- You should now see an RStudio environment with the required
dependencies pre-installed and loaded.