teenyverse

The Teenyverse. See project status for the ongoing development work.

Workflow

• Make sure you have quarto and julia installed on your machine.
• Clone this repo.
• You'll need to instantiate the virtual environment. From the command line, you can run:

  julia> ] 
  (@1.8) pkg> instantiate

  • Move to the directory of interests, e.g. teenyverse/posts/source-sink for the source-sink model.
• For each model, we have the following structure:

├── index.qmd              # Main file to render the interactive page. There should be only `ojs` cells in it.
├── models                 # All models live in their own directory
│   ├── sourcesink1.jl
│   ├── sourcesink2.jl
|   ├── ...
|   └── sourcesink_template.jl
├── sourcesink_sketch.jpg  # Sketch of the model
├── src
│   ├── 01_source-sinkDB.jl  # Add params to the SQLite DB
│   ├── 02_script2vacc.jl    # Script to generate bash scripts for params sweep on the VACC
│   └── 04_processing.jl     # Processing script before passing to index.qmd
├── source-sink.db         # SQLite DB containing grid of params for the models 
└── source-sink-res.db     # SQLite DB containing data to visualize

Trying out a new model

• In the model repo, there is a template file (e.g. teenyverse/posts/source-sink/models/sourcesink_template.jl)
• Make a copy of the template, numbering the file as the next iteration.
• Follow the script TODOs.
• Use the section at the bottom to protoype dynamically your file.

Adding a parameter sweep

• Update manually the parameter database of the model, using src/01_source-sinkDB.jl
• Run src/02_script2vacc.jl to create the bash scripts we will run in parallel on the vacc, e.g.

  julia src/02_script2vacc.jl --db "source-sink.db" -m "sourcesink2" -b 30

• Send everything though the vacc via the github repo.
• Move in the model directory, then run the following command:

  for file in $(ls sourcesink2_output/vacc_script/*.sh); do sbatch $file; done;

• Process the data, e.g.

  julia src/04_processing.jl -d "sourcesink2_output"

• Push back to github the source-sink-res.db, the dashboard should update itself.
• Backup the raw data somewhere.

Adding a new parameter configurations to an existing model

• Have the model rendered somewhere, either directly on the blog page or knitting the .qmd document (on VScode, this is ctrl-shift k when you are in the index.qmd file).

• Run new configurations as needed from the root dir of the blog page (e.g. source-sink/). For example, if you want to add beta values 0.27, gamma values 1.1., and the rest is the default, you can do:

  # the rest of the params will be the default below
  julia models/source-sink1.jl --beta 0.27 -g 1.1 

  Running julia models/source-sink1.jl --help will give you the argument names and current default values:

  
  usage: sourcesink1.jl [--db DB] [-L L] [-O O] [--beta BETA] [-g G]
                    [-r R] [-b B] [-c C] [-m M] [-o O] [-h]

optional arguments: --db DB Use Database to query parameters -L L LIMIT of rows (type: Int64, default: 5) -O O The OFFSET clause after LIMIT specifies how many rows to skip at the beginning of the result set. (type: Int64, default: 0) --beta BETA Spreading rate from non-adopter to adopter beta (type: Float64, default: 0.07) -g G Recovery rate gamma, i.e. rate at which adopters loose behavioral trait (type: Float64, default: 1.0) -r R Global behavioral diffusion rho (allows the behaviour to spread between groups) (type: Float64, default: 0.1) -b B Group benefits b (type: Float64, default: 0.18) -c C Institutional cost c (type: Float64, default: 1.05) -m M Noise u (type: Float64, default: 0.0001) -o O Output file for results (default: ".") -h, --help show this help message and exit

 Once this is done running, the rendered file should reflect the changes. To see the changes on the web page, we need to push the changes on Github. Then Github action will take care of update the web page.

### Modify/add new models

 - To modify an existing model, create a new file in `models` (see `source-sink/models/source-sink2.jl` for a template). Write the model in a function of the same name. This model should be added as a new tab in the `output` section in the main `index.qmd` file.
 - To add a new model create a new folder (see for example `teenyverse/posts/sci-group-life-cycle`). The folder should follow the same logic than first model.