Optimal Design of sero surveys

This repository contains the source code for the paper titled, "COVID-19: Optimal Design of Serosurveys for Disease Burden Estimation". The pre-print of the paper is available on arxiv.

The code is written mostly in Python with a few R code and have been used to get the results reported in Table 3. The required packages for the python scripts are mentioned under requirements.txt.

It is possible for users to set the total budget for the sero study (C), along with cost for each test namely, RAT, RT-PCR and iGG antibody tests. Note: The unit of all costs are in 1 thousand units

The python script is used for the results obtained from Theorem 1

python main.py -m local

which uses optimization/local.py to do an optimization for a given budget $C$ and a fixed parameter $p$ proposed in Theorem 1.

The worst case design can be obtained by running the script:

python main.py -m grid

which uses optimization/grid_search.py to get the optimization for the worst case-design proposed in Theorem 2.

In addition to the optimization methods, the main.py script also allows users to enter custom inputs using the following input flags:

input flag	description
-C	specify total budget
-cRAT	cost of one RAT test
-cRTPCR	cost of one RT-PCR Test
-cIGG	cost of one IGG test

The following snippet is an example for a complete input arguments supplied to the python script

python main.py -C 10000 -cRAT 0.5 -cRTPCR 0.1 -cIGG 0.45 -m grid

The R code for the optimization described in Theorem 1 is available in the R_code directory. The R code can be run for R studio, the path to the input files are set relative to the R_code directory.

Citation

If you are using this code, please consider citing the paper

@misc{athreya2020covid19,
      title={COVID-19: Optimal Design of Serosurveys for Disease Burden Estimation}, 
      author={Siva Athreya and Giridhara R. Babu and Aniruddha Iyer and Mohammed Minhaas B. S. and Nihesh Rathod and Sharad Shriram and Rajesh Sundaresan and Nidhin Koshy Vaidhiyan and Sarath Yasodharan},
      year={2020},
      eprint={2012.12135},
      archivePrefix={arXiv},
      primaryClass={stat.AP}
}

License

The source code is publicly available under the Apache2 license terms. Copyrights for the work belongs to the Indian Institute of Science Bangalore, Indian Statistical Institute Bangalore Centre, Indian Institute of Public Health Bangalore and Strand Life Sciences, Bangalore.