see19 0.3.5

An interface for visualizing and analysing the see19 dataset

see19

An aggregation dataset and interface for visualizing and analyzing Coronavirus Disease 2019 aka COVID19 aka C19

Dataset Last Updated June 12, 2020

---

May 31, 2020 Update

Upgrade to version 0.3 is complete. Please exercise caution if switching to this version as there have been a number of significant changes / additions that might impact your prior work.

SUMMARY OF UPDATES

1. Testset Graduation

• Test counts and Apple mobility data have been moved into the main dataset.
  • Reporting on testing continues to be inconsistent around the world. Many countries have only just begun reporting and many report on an infrequent basis (weekly or worse). Where there are gaps in daily figures, non-linear interpolation is used to smooth figures. Several key regions including Brazil and France have very minimal data at all.

2. Added filter functionality
When instanting a CaseStudy instance:

• You can now pass any of region_id, region_code, or region_name to regions/exclude_regions in a single iterable. region_code column has been added, and is either simply a replica of country_code or the accepted abbreviation of the province or state. i.e. Alberta's region_code is AB.
• country_code and country_id now also acceptable in countries/exclude_countries
• pandas Series and numpy arrays are now acceptable iterables for these filters as well.

3. Miscellaneous

• To access the testset via get_baseframe, set test=True
• Added progress bar for get_baseframe() (a couple hours I won't ever get back)
• Additional styling attributes to most chart make() functions
• Added exception to catch when a country_w_sub is provided as region when country_level=False
• when USA is filter via countries, see19 now automatically excludes the country of Georgia. This was a major personal irritant of mine, but if you have the need you can simply include Georgia in countries as well.

---

Latest Analysis

How Effective Is Social Distancing?

What Factors Are Correlated With COVID19 Fatality Rates?

The COVID Dragons

---

The Dataset

The dataset is in csv format and can be found here

You can find relevant statistics and detailed sourcing in the Guide

The Package

the see19 package is available on pypi and can be installed as follows:

pip install see19

The package provides a helpful pandas-based interface for working with the dataset. It also provides several visualization tools

The Guide

The Guide details data sources, structure, functionality, and visualization tools.

---

Purpose

"It is better to be vaguely right than exactly wrong."

- Carveth Read, Logic, Chapter 22

see19 is an early stage attempt to aggregate various data sources and analyze their impact (together and in isolation) on the virulence of SARS-CoV2.

• Ease-of-use is paramount, thus, all data from all sources have been compiled into a single structure, readily consumed and manipulated in the ubiquitous csv format.

see19 aggregates the following data:

• COVID19 Data Characteristics:
  • Cumulative Case, Fatality, and Testing statistics for each region on each date
  • State / Provincial-level data available for
• Factor Data Characteristics available for most regions include:
  • Longitude / Latitude, Population, Demographic Segmentation, Density
  • Climate Characteristics including temperatue and uvb radiation
  • Historical Health Outcomes
  • Travel Popularity
  • Social Distancing Implementation
  • And more and counting ...

There is no single all-encompassing data from an undoubted source that will serve the needs of every user for every use case. Thus, the dataset as it stands is an ad-hoc aggregation from multiple sources with eyeball-style approximations used in some instances. But while the dataset's imperfections are numerous, they cannot blunt the power of the insights that can be gleaned from an early exploratory analysis.

In addition to the dataset, see19 is a python package that provides:

• Helpful pandas-based interface for manipulating the data
• Visualization tools in bokeh and matplotlib to compare factors across multiple dimensions ..
• Statistical analysis is also a goal of the project and I expect to add such analysis tools as time progresses. Until then, the data is available for all.

---

Suggestions For Additional Data

I am always on the hunt for new additions to the dataset. If you have any suggestions, please contact me. Specifically, if you are aware of any datasets that might integrate nicely with see19 in the following realms:

1. German daily, state-level counts
2. Russian daily, state-level counts
3. India daily, sate-level counts
4. State or city level travel data
5. Global Commercial Airline route data (there seems to be plenty available, except only for a whopping price)

---

Quick Demo

You can very quickly use see19 to develop visuals for COVID19 analysis and presentation.

The see19 package can be installed via pip.

pip install see19

Then simply:

# Required to use Bokeh with Jupyter notebooks
from bokeh.io import output_notebook, show
output_notebook()

Loading BokehJS ...

from see19 import get_baseframe, CaseStudy
baseframe = get_baseframe()

regions = ['Germany', 'Spain']
casestudy = CaseStudy(baseframe, regions=regions, count_categories='deaths_new_dma_per_1M')

label_offsets = {'Germany': {'x_offset': 8, 'y_offset': 8}, 'Spain': {'x_offset': 5, 'y_offset': 5}}  
p = casestudy.comp_chart.make(comp_type='multiline', label_offsets=label_offsets, width=750)

show(p)

%matplotlib inline

regions = list(baseframe[baseframe['country'] == 'Brazil'] \
    .sort_values(by='population', ascending=False) \
    .region_name.unique())[:20]

casestudy = CaseStudy(
    baseframe, count_dma=5, 
    factors=['temp'],
    regions=regions, start_hurdle=10, start_factor='cases', lognat=True,
)
kwargs = {
    'color_factor': 'temp',
    'fs_xticks': 16, 'fs_yticks': 12, 'fs_zticks': 12,
    'fs_xlabel': 12, 'fs_ylabel': 18, 'fs_zlabel': 18,
    'title': 'Daily Deaths in Brazil as of May 2',
    'x_title': 0.499, 'y_title': 0.738, 'fs_title': 22, 'rot_title': -9.5,
    'x_colorbar': 0.09, 'y_colorbar': .225, 'h_colorbar': 20, 'w_colorbar': .01, 
    'a_colorbar': 'vertical', 'cb_labelpad': -57,
    'tight': True, 'abbreviate': 'first', 'comp_size': 10,
}
p = casestudy.comp_chart4d.make(comp_category='deaths_new_dma_per_1M', **kwargs)