Data Science in Global Health: Programming Library/Resource List

Author(s): MEASURE Evaluation

Year: 2018

Abstract:

Getting started with data science can be a daunting task. Identifying and understanding the appropriate methods and techniques presents one challenge, while producing the code necessary to implement those methods and techniques presents another challenge.

MEASURE Evaluation’s Programming Library/Resource list seeks to address those twin challenges by providing links to existing resources that present an overview of data science techniques and example code for applying those techniques. The links in this resource list are intended for global health professionals with basic to moderate programming skills and who are interested in strengthening their knowledge and experience in data science methods and techniques.

Links with R code

UCLA Resources
This site provides multiple links to R resources including links for downloading software: https://stats.idre.ucla.edu/r/

The site includes links to free online modules introducing fundamentals of R including importing data, variable construction, and data visual representation: 
https://stats.idre.ucla.edu/r/modules/

Additional links provide code for data visual representation and various statistical methods including descriptive statistics, logistic regression, and multilevel models: https://stats.idre.ucla.edu/r/codefragments/introduction/ and 
https://stats.idre.ucla.edu/other/dae/

UNC Resources
Open course materials presented as separate lectures for various statistical methods, including multiple negative binomial regression. The site also includes lectures introducing Bayesian methods for regression analysis, and graphical representations in R. Each lecture page provides sample R code for each method presented: 
https://www.unc.edu/courses/2010fall/ecol/563/001/docs/lectures/lecture1.htm

Rspatial
This site provides code and descriptions for spatial data analysis in R, as well as links to download various spatial analysis packages. The site also includes an introduction to manipulating spatial data in R: http://www.rspatial.org/

Code Index
Collection of code snippets for specific tasks: https://source.opennews.org/code/

Starting data analysis/wrangling with R: Things I wish I had been told
Descriptions of data wrangling techniques and associated R code: 
http://reganmian.net/blog/2014/10/14/starting-data-analysiswrangling-with-r-things-i-wish-id-been-told/

R Reference Card for Data Mining
List of R packages and functions that can assist with data mining: 
http://www.rdatamining.com/docs/r-reference-card-for-data-mining

Causal Impact: Estimating causal effects in time series
Open-source R package developed by Google that makes causal analyses simple and fast. based on Bayesian structural time-series models. We use these models to construct a synthetic control—what would have happened to our outcome metric in the absence of the intervention. This approach makes it possible to estimate the causal effect that can be attributed to the intervention, as well as its evolution over time: https://opensource.googleblog.com/2014/09/causalimpact-new-open-source-package.html

RAPPOR
R and Python open source project from Google to facilitate analysis of data while preserving privacy of individuals: 
https://github.com/google/rappor and https://ai.googleblog.com/2014/10/learning-statistics-with-privacy-aided.html

Mapping with ggplot: Create a nice choropleth map in R
Basic tutorial on how to produce choropleth maps in R: 
http://rforpublichealth.blogspot.com/2015/10/mapping-with-ggplot-create-nice.html

Introduction to R Markdown (R Notebooks)
R Markdown is the R version of Jupyter notebooks. Code can be written and executed in a document that makes it possible to present findings and the output of analysis but also include the code used to conduct analysis in a way that makes reproducible results possible: https://rmarkdown.rstudio.com/r_notebooks.html

Flexdashboard: Easy interactive dashboards for R
Interactive dashboards produced using R Markdown. Can be linked with Shiny to make visualizations dynamic: https://rmarkdown.rstudio.com/flexdashboard/

Implementation of a reproducible data analysis workflow
Overview of best practice for creating a workflow in R that can be replicated. Replicable workflows make it possible to execute analysis of data by launching one R script and can facilitate collaboration: 
http://blog.jom.link/implementation_basic_reproductible_workflow.html

Access WHO Global Health Observatory Data from R
Github repository of code to access data from the WHO Global Health Observatory within R: https://github.com/pierucci/rgho and https://cran.r-project.org/web/packages/WHO/vignettes/who_vignette.html

Obtaining DHS Data via API
Code snippets for access to DHS data via an API: 
https://api.dhsprogram.com/#/samples-python.cfm and http://api.dhsprogram.com/#/samples-r.cfm

Apps to visualize or map data: http://api.dhsprogram.com/#/sample-apps.cfm

DHS.rates R package: https://cran.r-project.org/web/packages/
DHS.rates/vignettes/DHS.rates.html

Links with Python code

Python for data science
This short primer on Python is designed to provide a rapid “on-ramp” to enable computer programmers who are already familiar with concepts and constructs in other programming languages learn enough about Python to facilitate the effective to use of open-source and proprietary Python-based machine learning and data science tools:  
http://nbviewer.jupyter.org/github/gumption/Python_for_Data_Science/blob/
master/Python_for_Data_Science_all.ipynb

Python basics for data science
Tutorial on using Python and Jupyter notebooks. Provides an overview of key Python libraries and techniques and an introduction to Jupyter notebooks: 
https://data36.com/python-for-data-science-python-basics-1/

Bayesian First Aid: Pearson Correlation Test
Understanding the degree of correlation between two variables is a key step in effective use of data for decisions. Bayesian analysis offers the opportunity to assess correlation in data that doesn’t have a normal distribution: 
http://www.sumsar.net/blog/2014/03/bayesian-first-aid-pearson-correlation-test/

General Data Science Links

The Data Science Toolbox is a virtual environment based on Ubuntu Linux that is specifically suited for doing data science. Its purpose is to get you started in a matter of minutes. You can run the Data Science Toolbox either locally (using VirtualBox and Vagrant) or in the cloud: http://datasciencetoolbox.org/

Cross Validated is a question and answer site for people interested in statistics, machine learning, data analysis, data mining, and data visualization: 
https://stats.stackexchange.com/