wwaites / stamford-hill

Scripts, data processing and models for Stamford Hill

Data processing and modelling for Stamford Hill

Introduction

This repository contains scripts and programs for processing data collected as part of the project to understand the spread of COVID-19 in a Haredi Jewish community in North London. There are two primary sources of data: information about households and locations that they frequent collected by survey and serology data. It is not possible to distribute this data for confidentiality reasons.

The programs are therefore separated into four groups:

• Group 1 contains programs that produce, from the source data, anonymous statistics and other representations that can be distributed. These data are used as input to the Group 2 programs. These are:
  • stamford_graph produces documents in GraphML format from the survey data. The attributes representing demographic and other information on the nodes and edges are minimised and pseudonymous identifiers for individuals, households and places are used.
  • stamford_plot produces basic demographic plots and summary information about the sampled population.
• Group 2 contains programs that must not use any identifiable information as input. These programs take the statistical and other representations of the input data about the sampled population and produce synthetic data representative of the entire population.
• Group 3 contains programs for simulations of transmission in the community.
• Group 4 contains programs for processing the output of simulations into summaries, plots and so forth for including in publications and other materials.

Installation

This repository is not intended to be specific to any particular programming language. We use a mix of Python, R and other languages as convenient. The programs are intended to be run as command-line executables. This, however, requires some language-specific setup.

Python

The easiest is to use anaconda or similar software for virtual environments. We require Python 3 throughout. A typical installation for development purposes is done as follows:

conda create -n stamford python=3
conda activate stamford
conda install numba networkx scipy matplotlib pandas
python setup.py develop

now the programs written in python will be in your shell's search path and you will simply be able to run them.

Programs

stamford_graph

usage: stamford_graph [-h] [--maximal] [--motifs] survey

positional arguments:
  survey        Data file (.zip) containing survey data downloaded from ODK

optional arguments:
  -h, --help    show this help message and exit
  --maximal     Do not minimise output; store all annotations in the graph
  --motifs, -m  Generate a graph of household motifs

Run this program like this:

stamford_graph ./data/survey.zip > stamford.graphml
stamford_graph -m ./data/survey.zip > stamford-motifs.graphml

The first form creates a graph of the entire community. The second produces household motifs for each household pattern that appears more than once in the data.

stamford_plot

usage: stamford_plot [-h] graph

positional arguments:
  graph       Sampled population graph

optional arguments:
  -h, --help  show this help message and exit

This program writes histograms for the numbers of households, shuls, yeshivas and mikvas that associated with a given number of people.

stamford_house

usage: stamford_house [-h] [--nrestarts NRESTARTS] [--seed SEED] [--lower LOWER] [--upper UPPER] graph

positional arguments:
  graph                 GraphML file containing network

optional arguments:
  -h, --help            show this help message and exit
  --nrestarts NRESTARTS
                        Number of times to run the fitting process
  --seed SEED           Random seed
  --lower LOWER         Lower age cutoff
  --upper UPPER         Upper age cutoff

This program fits a household transmission model to the cases present in the graph. It estimates the baseline transmission probability in households as well as the degree of over- or under-influence of a particular age band.