Posts for Tag: data

We can compare the sizes of countries and continents to planets and moons by projecting a map of a specific country onto another planet. Select a country and planet or moon to find out.

In one of my kid’s favorite books, there’s a picture demonstrating how Pluto is the same size as Australia. It has a satellite image of the country and an image of the former ninth planet superimposed on top as if it were hovering above the country. That image has stuck with me and I thought it would be interesting to see how other countries would compare with other planets and bodies in our solar system. As I’ve been working with javascript graphing/mapping library, D3.js and making maps/globes, I realized I should try to “project” individual countries onto these planets to see what they looked like.

Instructions

This visualization should be pretty self explanatory. You can select a country or continent and a planet or moon (or the sun) in the solar system. The visualization will then project the land onto the body and you have a simple visual comparison of the size of the country/continent and the planet or moon. You can drag on the visualization to rotate the planet.

There are some combinations that are not possible because the country/continent is too large to be projected onto the body without overlap. In these cases, the planet or country will be greyed out in the selection menu. You can click the “Get URL” button and share a specific map combination (country and planet) by copying the address in the url address bar.

The visualization also displays the area of the country/continent and the surface area of the planet or body. In some cases, the percentage may not look correct but remember that you can only see half of the planet surface and that it’s actually a hemisphere (half a sphere and not just a circle). It becomes clearer if you draw the surface of the planet around.

Calculations

The calculations to project a country onto another body involves starting with a set of coordinates (made up of longitude and latitude values) which define the border of the country, in the geojson format. To display them on Earth, the coordinates are modified so that the center of the country is centered at the intersection between the equator and prime meridian [0 deg latitude, 0 deg longitude].

To display them projected on a different planet or moon, it is necessary to change the latitude and longitude values of each point of the polygon country border so that it represents the same distance away from the polygon center. I used the Haversine formula to calculate the distance and bearing between two points on a sphere and then used the inverse to find the coordinates that were that distance and bearing from the center point on a sphere of a different size. These formulas can be found here. The main idea is that the distance representing one degree of latitude on Earth will be half as large on a planet that is half the size of Earth (like Mars). Thus, the distance between the center of a country and a point on the border will be a different number of degrees latitude and longitude from the center point on a different planet than on Earth. And this calculatin is done using these formulae.

Sources and Tools:
This visualization was made using the open-source, d3 javascript dataviz library and UI are made using HTML, CSS and javascript.

Visualize the large number of coronavirus cases and deaths in the US each day/hour in about 10 seconds

The rate of COVID-19 deaths and cases in the US is crazy high after the 2020 winter holidays and maybe still be going up. This visualization shows the number of COVID cases that occur in one hour or the COVID deaths that occur in one day based on the average of the last five days. This is another attempt to show the true scale of how many cases and deaths the US is dealing with, since it is often hard to understand large numbers. I have also attempted to show the scale of US deaths/cases here and here. Unfortunately, there are so many people getting sick and dying, it’s hard to fathom just how many people this actually is.

The 5-day averaging was done to smooth out any peaks and troughs in data reporting due to weekends/holidays, since I noticed that some states were literally reporting zero COVID cases some days while reporting many hundreds or thousands of cases other days.

The dots shown on the animation are located in the state that the cases or deaths occur but are randomly spread out within the state. This is done for visual clarity since if they were shown in their actual location, most of the dots would be overlapping in urban, high density areas. This approach lets you see which states have high COVID instances but still locate them by state.

You can share this animation by putting ?cat=deaths or ?cat=cases behind the url or copying and sharing one of these links:

Sources and Tools:

The coronavirus data comes from the covidtracking.com API. The data is parsed daily using a custom python script and visualizations are made using the open-source Leaflet javascript mapping library and the interface and animation are made using HTML/CSS/javascript.

How do 2020 presidential election results correlate with population density?

The visualization I made about county election results and comparing land area to population size was very popular around the time of the 2020 presidential election. As the counties were represented by population, it was clear that democratic-leaning areas on that map tended to grow in size, while republican-leaning areas tended to shrink. This raised the question of exactly how population density correlates with election results.

Hover over (or click on) the bubbles to see information about the county.

It’s clear there is a very strong correlation between the vote margin and population density.  Vote margin is the percentage amount that one candidate beat the other candidate by in the county (0% means a tie while 50% means that one candidate got 75% and the other got 25% of the voteshare).  Population density is calculated as people per square mile in the county and is shown in the graph on a log scale, where each major grid line is 10 time greater than the previous one.  This is done because there is one to two orders of magnitude difference in the densest counties (in New York City) and even moderately dense counties.  There are also several counties with population density below 1 person per square mile (several in Alaska because of the size of their counties) but these are excluded from the graph.

Richmond County, NY (i.e. the Borough of Staten Island) is the densest county (17th densest) in the country that Trump won. The densest counties favored Biden quite heavily as he won 45 of the 50 densest counties in the country, which also tend to have a fairly high population.

This second graph is a histogram that specifically categorizes counties into discreet bins by population density. Note that they are on a log scale as well. You can toggle the graph to show the number of counties won by each candidate or the number of votes won in each of the population density bins. The black line shows the percentage of counties (or votes) won by the democratic candidate (Joe Biden) in each of those bins.

Hover over (or click on) the bars to see information about each county bin.

It’s pretty clear in these graphs that low population density areas clearly favor the republican while the denser areas favor the democrat.

Data and Tools
The 2020 county-level election data is downloaded from the New York Times county election data API and processed using a python script. Population data used is for 2018. The visualization was created using the open-source plotly javascript graphing library.

The number of US coronavirus cases is equal to the population of several states put together.

Sources and Tools:

Each state has two senators in the Senate, even though there is a great disparity in the populations of the states. This was a compromise that the framers of the Constitution dealt with in creating the framework of the US government. While the US House of Representatives is based on proportional representation, the Senate was designed to have two senators per state regardless of population. This leads to some interesting variations in the number of votes that some senators get relative to other senators (and how many people they represent).

Graph of Total Votes for Each Current Senator (2014, 2016 and 2018)

This graph is called a treemap and shows the total number of votes cast for the winner of each senate race of the current sitting senators. They are shown in order from largest to smallest vote totals, where the area of the rectangle is proportional to the number of votes. The treemap can be organized by party if desired. This graph does not show the number of votes that their opponents got.
If you hover over (click, on mobile) one of the boxes in the treemap, you can compare the number of votes received by that senator to the number of senators that received the same number of votes combined. This helps highlight the disparities in the representation of voters in large states in the Senate relative to that of voters in states with low populations.

For example, Kamala Harris, Democratic senator of my home state of California, received 7.5 million votes when she won her senate race in 2016. This large number of votes is larger than the combined votes for 22 of her Republican colleagues in small states. This is even more impressive since, as noted before, she ran against another Democrat Loretta Sanchez, in the election.

Note that some of the recently elected senators shown in the table are no longer serving in the Senate:

• John McCain’s seat is currently held by Martha McSally
• Johnny Isakson’s seat is currently held by Kelly Loeffler

Because of the large variation in population sizes and a tendency for more populous states to vote for democrats, Democratic Senators received many more votes in their elections than their Republican colleagues did, despite having fewer numbers. The 47 Democratic (and Independent) senators received a total of 67.5 million votes while the 53 Republican senators received 59.5 million votes.

Graph of Margin of Victory over Opposing Party for Each Current Senator (2014, 2016 and 2018)

This graph shows a slightly different set of data. Instead of total votes for the winning candidate, it shows the vote margin (i.e. the number of votes the winner received vs the opponent of a different party). The reason I specify it this way is that the two Democratic California senators defeated other democrats to win their elections (i.e. no republican was on the ballot in the general election because no republican got enough votes in the primary). This comparison is interesting because not only do some senators receive very few votes (because they live in small states), but they may only win by a small margin over their opponents. Comparing margins of victory, shows how few votes it would take to “flip” a Senate seat between the two parties.

If you take Kamala Harris’s margin of victory over Republicans to be her vote total (7.5 million votes) since there was no Republican running against her, her margin of victory is greater than the margin of victory of 43 of her Republican Senate colleagues combined.

Sources and Tools:

Senate election data was downloaded from MIT election lab. The data was processed using python/pandas and the visualization was created using javascript and plotly.js, the open source javascript graphing library.

The US coronavirus death rate is quite high compared to other countries (on a population-corrected basis)

US coronavirus deaths have surpassed 300,000. Many of these deaths could have been avoided if swift action had been taken in February and March, as many other countries did. This graph shows an rough estimate of the number of US deaths that could have been avoided if the US had acted similar to other countries.

This graph takes the rate of coronavirus deaths by country (normalized to their population size) and imagines what would happen if the US had had that death rate, instead of its own. It then applies that reduction (or increase) in death rate to the total number of deaths that the US has experienced. The US death rate is about 600/million people in September 2020 and if a country has a death rate of 60/million people, then 90% of US deaths (about 180,000 people) could have been avoided if the US had matched their death rate. The government response to the pandemic is one of several important factors that determine the number of cases and deaths in a country. This means proper messaging about the need to wear masks and socially distance as well as providing payments to citizens and business to help them during the economic shutdown. Other important factors can include the overall health of the population, the population structure (i.e. age distribution of population), ease of controlling borders to prevent cases from entering the country, presence of universal or low-cost health care system, and relative wealth and education of the population.

The graph lets you compare the potential reduction in US deaths when looking at 30 different countries. You can choose those 30 countries based on total population, GDP or GDP per capita. These give somewhat different sets of countries to compare death rates, which is an indication of the effectiveness of the coronavirus response.

A valid criticism of this graph is that testing and data collection is very different in each of the countries shown and the comparisons are not always valid. This is definitely a problem with all coronavirus data but for the most part, the very large differences between death rates would still exist even if data collection were totally standardized. Some of the data from the poorest countries is less reliable, because they have less testing capabilities.

Source and Tools:
Data on coronavirus deaths by country is from covid19api.com and downloaded and cleaned with a python script. Graph is made using the plotly open source javascript library.