Mini wrapper for dismod_at

Wrapper(s) that fits a two-level nonlinear ODE-constrained random effects model using dismod_at. Also includes code files for related utilities functions (e.g. plotting, matching measurement data with covariates).

dismod_at

Documentation can be found here.

Install and Run dismod_at

• dismod_at can be installed locally via a docker image. Change the hash on the git checkout line in the Dockerfile to that of the latest commit
• Build a dismod_at image using

docker build -t dismod_at:{tag} {path_to_docker_file}

• Use docker images to check that the image is built successfully.

• To run the demo jupyter notebooks, use the command

  docker run -v "$(pwd)":/home/work -p 8890:8890 -it dismod_at:{tag}
  

  You can change the host port number.

• Inside a docker container, run

  jupyter notebook
  

  Copy the link popped out in the terminal and open it in the browser. You can edit

  /root/.jupyter/jupyter_notebook_config.py
  

  to change settings.

Wrapper files

• dismod_db.py: create a dismod_at database.
• two_level.py: run a two_level nonlinear random effects model (i.e. one parent with multiple children). also has the option to run just one-level, ignoring the parent-children structure.
• plot_two_level.py: plot residuals, age/time pattern for rates and covariate values for each child.
• dismod_output.py: generate output that conforms to GBD format at IHME.
• covariate.py: match covariate to measurement data using population weighted smoothing.

older

• fixed_only.py: run a nonlinear regression ignoring the parent-children structure in the data.
• plot_fixed_only.py: plot residuals, age/time pattern for rates and covariate values.