rbpatt2019 / dash-covid19

An interactive Dash app for viewing the WHO's Covid-19 data

dash-covid19: A Dash App for Covid-19 Data

Introduction

dash-covid19 is a Dash app for viewing up-to-date information related to the spread of Covid-19 using the data provided by a variety of sources collated by Our World in Data.

Docs and Code

This app is hosted by Heroku and lives here.

The helpful hints are embedded throughout the app. Any where you see underlined text, you can mouse over to get a pop-up about what each component does.

The documentation lines here.

The code lives in here.

Usage

The app contains 5 pages and 3 external links. The home page provides a brief summary of all the assorted links within the app, as well as an explanation of why I built the app.

The "Overview" page provides a concise display that summarises the impact of Covid-19 in a country using population adjusted statistics.

The "Explorer" page provides an interactive graph to help the user, well, explore the most up-to-date Covid-19 data. This page is constantly changing, so check back to see how the graphs evolve!

The "World Map" pages provides an interactive, global map that plots information on top of, well, a map! Here you can see not only interactions, but regional impacts as well. You can change the data represented by the size and colour of the dots.

The "Data Table" pages displays the data in an interactive table that allows the user to sort and filter columns of interest. The four left-most columns (iso_code, continent, location, and date) remain fixed as you scroll.

As this project is proudly open source, the Navbar also includes three external links. These will take you back to this repo to view the code, to OWID's repo to view the data, or to ReadTheDocs to view the source code documentation. This way, you can see exactly where the data comes from, and how it is being handled.

About the Data

Fundametally, good data science requires two things: good data and good questions. Since good questions frequently evolve from exploratory data analysis, one could argue that good data science needs only one things: good data. Crap in, crap out, as they say... Now, how does one get data? As a trained biologist, my first thought is structured experiments, with repeated samples, controls groups, and as many data points as feasibly possible. Unfortunately, this structured framework isn't always possible, particularly when dealing with rpaidly evolving, as-of-yet not fully understood situations like the current Covid-19 pandemic. In these cases, where limitations are inherent to the problem, transparency becomes critical. We must accept that the data are not, and cannot be, perfect, recognise the limitations in the data, and be open about any limitations when discussing interpretations of the data.

To that end, I have elected to use data collated by the team at Our World in Data (OWID) when constructing my app. The data are freely avaialble, and can be found in their repo. Their principle source of data is the European Center for Disease Prevention and Control (ECDC), who provide daily updates of worldwide statistics. OWID thoroughly discusses all their sources, methods for data handling, and the reasoning behind any decisions they make relating to data management. Please, visit their repo for any questions you may have about the data.

Recent Changes

Please see the CHANGELOG

Contributing

If you would like to contribute to development, please see the instructions <CONTRIBUTING.md>

Next Steps

• Update data summary colours based on 7 day averages
• Continue to expand documentation
• Caching to improve performance

Thank Yous

• Dash for developing an incredible framework and excellent documentation.
• Dash Bootstrap Components for making it so much easier to layout the app.
• OWID for providing an incredible resource amid a rapidly evolving situation.

dash-covid19's Issues

Negative values in data prevent graph rendering

Some columns (new_cases, new_deaths were where it is noticed) containing a small number of negative values. This causes the graph to fail to render with no obvious error.

Solution: crop values to 0 for plot, but not for data table.

Realistically, I just need to migrate to an SQL database.

Travis build fails on staging branch with 'percy has no attribute Runner'

Builds to travis fail with the full error:

E           AttributeError: module 'percy' has no attribute 'Runner'

Confusingly, tests with Percy work exactly as expected when run from the command line with make test. I suspect it's something to do with the way Travis is handling Percy installs.

Deploys to Heroku fail

Deploying to Heroku with tags from the master branch fails because You do not have access to the app dash-covid19.

Pylint fails with E1102 on dash.daq components

Raised on the dash.daq repo.

Travis build fails as DashComposite is note defined

Think this results from testing extra not getting installed.

Makefile is ignoring variable expansion

make patch,minor, major all fail to work as expected as variable expansion is not occurring in the command

select_dcc_dropwdown fails when located beneath reactive graph

I'm using dash_duo.select_dcc_dropwdownto test callbacks related to the dropdown. Interestingly, the test fails if the dropdown is located below the reactive graph, as the element cannot be found. However, if it is located above the graph, or if there is no graph at all, the tests pass as expected.