Ranking data provide context but can also confuse

This dataviz from the Economist had me spending a lot of time clicking around - which means it is a success.

The graphic presents four measures of wellbeing in society - life expectancy, infant mortality rate, murder rate and prison population. The primary goal is to compare nations across those metrics. The focus is on comparing how certain nations (or subgroups) rank against each other, as indicated by the relative vertical position.

The Economist staff has a particular story to tell about racial division in the US. The dotted bars represent the U.S. average. The colored bars are the averages for Hispanic, white and black Americans. The wider the gap between the colored bars, the more variant is the experiences between American races.

The chart shows that the racial gap of life expectancy is the widest. For prison population, the U.S. and its racial subgroups occupy many of the lowest (i.e. least desirable) ranks, with the smallest gap in ranking.

***

The primary element of interactivity is hovering on a bar, which then highlights the four bars corresponding to the particular nation selected. Here is the picture for Thailand:

According to this view of the world, Thailand is a close cousin of the U.S. On each metric, the Thai value clings pretty near the U.S. average and sits within the range by racial groups. I'm surprised to learn that the prison population in Thailand is among the highest in the world.

Unfortunately, this chart form doesn't facilitate comparing Thailand to a country other than the U.S as one can highlight only one country at a time.

***

While the main focus of the chart is on relative comparison through ranking, the reader can extract absolute difference by reading the lengths of the bars.

This is a close-up of the bottom of the prison population metric:

The length of each bar displays the numeric data. The red line is an outlier in this dataset. Black Americans suffer an incarceration rate that is almost three times the national average. Even white Americans (blue line) is imprisoned at a rate higher than most countries around the world.

As noted above, the prison population metric exhibits the smallest gap between racial subgroups. This chart is a great example of why ranking data frequently hide important information. The small gap in ranking masks the extraordinary absolute difference in incareration rates between white and black America.

The difference between rank #1 and rank #2 is enormous.

The opposite situation appears for life expectancy. The life expectancy values are bunched up especially at the top of the scale. The absolute difference between Hispanic and black America is 82 - 75 = 7 years, which looks small because the axis starts at zero. On a ranking scale, Hispanic is roughly in the top 15% while black America is just above the median. The relative difference is huge.

For life expectancy, ranking conveys the view that even a 7-year difference is a big deal because the countries are tightly bunched together. For prison population, ranking shows the view that a multiple fold difference is "unimportant" because a 20-0 blowout and a 10-0 blowout are both heavy defeats.

***

Whenever you transform numeric data to ranks, remember that you are artificially treating the gap between each value and the next value as a constant, even when the underlying numeric gaps show wide variance.

 

 

 

 

 

Hanging things on your charts

The Financial Times published the following chart that shows the rollout of vaccines in the U.K.

(I can't find the online link to the article. The article is titled "AstraZeneca and Oxford face setbacks and success as battle enters next phase", May 29/30 2021.)

This chart form is known as a "streamgraph", and it is a stacked area chart in disguise. 

The same trick can be applied to a column chart. See the "hanging" column chart below:

The two charts show exactly the same data. The left one roots the columns at the bottom. The right one aligns the middle of the columns. 

I have rarely found these hanging charts useful. The realignment makes it harder to compare the sizes of the different column segments. On the normal stacked column chart, the yellow segments are the easiest to compare because they share the same base level. Even this is taken away from the reader on the right side.

Note also that the hanging version does not admit a vertical axis. 

The same comments apply to the streamgraph.

***

Nevertheless, I was surprised that the FT chart shown above actually works. The main message I learned was that initially U.K. primarily rolled out AstraZeneca and, to a lesser extent, Pfizer, shots while later, they introduced other vaccines, including Johnson & Johnson, Novavax, CureVac, Moderna, and "Other". 

I can also see that the supply of AstraZeneca has not changed much through the entire time window. Pfizer has grown to roughly the same scale as AstraZeneca. Moderna remains a small fraction of total shots. 

I can even roughly see that the total number of vaccinations has grown about six times from start to finish. 

That's quite a lot for one chart, so job well done!

There is one problem with the FT chart. It should have labelled end of May as "today". Half the chart is history, and the other half is the future.

***

For those following Covid-19 news, the FT chart is informative in a different way.

There is a misleading statement going around blaming the U.K.'s recent surge in cases on the Astrazeneca vaccine, claiming that the U.K. mostly uses AZ. This chart shows that from the start, about a third of the shots administered in the U.K. are Pfizer, and Pfizer's share has been growing over time. 

U.K. compared to some countries mostly using mRNA vaccines

U.K. is almost back to the winter peak. That's because the U.K. is serious about counting cases. Look at the state of testing in these countries:

What's clear about the U.S. case count is that it is kept low by cutting the number of tests by two-thirds, thus, our data now is once again severely biased towards severe cases. 

We can do a back-of-the-envelope calculation. The drop in testing may directly lead to a proportional drop in reported cases, thus removing 500 (asymptomatic, or mild) cases per million from the case count. The case count goes below 250 per million so the additional 200 or so reduction is due to other reasons such as vaccinations.

Did prices go up or down? Depends on how one looks at the data

The U.S. media have been flooded with reports of runaway inflation recently, and it's refreshing to see a nice article in the Wall Street Journal that takes a second look at the data. Because as my readers know, raw data can be incredibly deceptive.

Inflation typically describes the change in price level relative to the prior year. The month-on-month change in price levels is a simple seasonal adjustment used to remove the effect of seasonality that masks the true change in price levels. (See this explainer of seasonal adjustment.)

As the pandemic enters the second year, this methodology is comparing 2021 price levels to pandemic-impacted price levels of 2020. This produces a very confusing picture. As the WSJ article explains, prices can be lower than they were in 2019 (pre-pandemic) and yet substantially higher than they were in 2020 (during the pandemic). This happens in industry sectors that were heavily affected by the economic shutdown, e.g. hotels, travel, entertainment.

Here is how they visualized this phenomenon. Amusingly, some algorithm estimated that it should take 5 minutes to read the entire article. It may take that much time to understand properly what this chart is showing.

Let me save you some time.

The chart shows monthly inflation rates of hotel price levels.

The pink horizontal stripes represent the official inflation numbers, which compare each month's hotel prices to those of a year prior. The most recent value for May of 2021 says hotel prices rose by 9% compared to May of 2020.

The blue horizontal stripes show an alternative calculation which compares each month's hotel prices to those of two years prior. Think of 2018-9 as "normal" years, pre-pandemic. Using this measure, we find that hotel prices for May of 2021 are about 4% lower than for May of 2019.

(This situation affects all of our economic statistics. We may see an expansion in employment levels from a year ago which still leaves us behind where we were before the pandemic.)

What confused me on the WSJ chart are the blocks of color. In a previous chart, the readers learn that solid colors mean inflation rose while diagonal lines mean inflation decreased. It turns out that these are month-over-month changes in inflation rates (notice that one end of the column for the previous month touches one end of the column of the next month).

The color patterns become the most dominant feature of this chart, and yet the month-over-month change in inflation rates isn't the crux of the story. The real star of the story should be the difference in inflation rates - for any given month - between two reference years.

***

In the following chart, I focus attention on the within-month, between-reference-years comparisons.

Because hotel prices dropped drastically during the pandemic, and have recovered quite well in recent months as the U.S. reopens the economy, the inflation rate of hotel prices is almost 10%. Nevertheless, the current price level is still 7% below the pre-pandemic level.

 

 

Circular areas offer misleading cues of their underlying data

John M. pointed me on Twitter to this chart about the progress of U.S.'s vaccination campaign:

This looks like a White House production, retweeted by WHO. John is unhappy about this nested bubble format, which I'll come back to later.

Let's zoom in on what matters:

An even bigger problem with this chart is the Q corner in our Trifecta Checkup. What is the question they are trying to address? It would appear to be the proportion of population that has "already received [one or more doses of] vaccine". And the big words tell us the answer is 8 percent.

But is that really the question? Check out the dark blue circle. It is labeled "population that has already received vaccine" and thus we infer this bubble represents 8 percent. Now look at the outer bubble. Its annotation is "new population that received vaccine since January 27, 2021". The only interpretation that makes sense is that 8 percent  is not the most current number. If that is the case, why would the headline highlight an older statistic, and not the most up-to-date one?

Perhaps the real question is how fast is the progress in vaccination. Perhaps it took weeks to get to the dark circle and then days to get beyond. In order to improve this data visualization, we must first decide what the question really is.

***

Now let's get to those nested bubbles. The bubble chart is a format that is not "sufficient," by which I mean the visual by itself does not convey the data without the help of aids such as labels. Try to answer the following questions:

In my view, if your answer to the last question is anything more than 5 seconds, the dataviz has failed. A successful data visualization should not make readers solve puzzles.

The first two questions depict the confusing nature of concentric circle diagrams. The first data point is coded to the inner circle. Where is the second data point? Is it encoded to the outer circle, or just the outer ring?

In either case, human brains are not trained to compare circular areas. For question 1, the outer circle is 70% larger than the smaller circle. For question 2, the ring is 70% of the area of the dark blue circle. If you're thinking those numbers seem unreasonable, I can tell you that was my first reaction too! So I made the following to convince myself that the calculation was correct:

Circular areas offer misleading visual cues, and should be used sparingly.

[P.S. 2/10/2021. In the next post, I sketch out an alternative dataviz for this dataset.]

Handling partial data on graphics

Last week, I posted on the book blog a piece about excess deaths and accelerated deaths (link). That whole piece is about how certain types of analysis have to be executed at certain moments of time.  The same analysis done at the wrong time yields the wrong conclusions.

Here is a good example of what I'm talking about. This is a graph of U.S. monthly deaths from Covid-19 during the entire pandemic. The chart is from the COVID Tracking Project, although I pulled it down from my Twitter feed.

There is nothing majorly wrong with this column chart (I'd remove the axis labels). But there is a big problem. Are we seeing a boomerang of deaths from November to December to January?

Not really. This trend is there only because the chart is generated on January 12. The last column contains 12 days while the prior two columns contain 30-31 days.

The Trifecta Checkup picks up this problem. What the visual is showing isn't what the data are saying. I'd call this a Type D chart.

***

What to fix this?

One solution is to present partial data for all the other columns, so that the readers can compare the January column to the others.

One critique of this is the potential seasonality. The first 38% (12 out of 31) of a month may not be comparable across months. A further seasonal adjustment makes this better - if we decide the benefits outweight the complexity.

Another solution is to project the full-month tally.

The critique here is the accuracy of the projection.

But the point is that not making the adjustment would be worse.

 

 

Dreamy Hawaii

I really enjoyed this visual story by ProPublica and Honolulu Star-Advertiser about the plight of beaches in Hawaii (link).

The story begins with a beautiful invitation:

This design reminds me of Vimeo's old home page. (It no longer looks like this today but this screenshot came from when I was the data guy there.) In both cases, the images are not static but moving.

The tour de force of this visual story is an annotated walk along the Lanikai Beach. Here is a snapshot at one of the stops:

This shows a particular homeowner who, according to documents, was permitted to rebuild a destroyed seawall even though officials were supposed to disallow reconstruction in order to protect beaches from eroding. The property is marked on the map above. The image inside the box is a gif showing waves smashing the seawall.

As the reader scrolls down, the image window runs through a carousel of gifs of houses along the beach. The images are synchronized to the reader's progress along the shore. The narrative makes stops at specific houses at which point a text box pops up to provide color commentary.

***

The erosion crisis is shown in this pair of maps.

There's some fancy work behind the scenes to patch together images, and estimate the boundaries of th beaches.

***

The following map is notable for its simplicity. There are no unnecessary details and labels. We don't need to know the name of every street or a specific restaurant. Removing excess details makes readers focus on the informative parts. 

Clicking on the dots brings up more details.

***

Enjoy the entire story here.

Convincing charts showing containment measures work

The disorganized nature of U.S.'s response to the coronavirus pandemic has created a sort of natural experiment that allows data journalists to explore important scientific questions, such as the impact of containment measures on cases and hospitalizations. This New York Times article represents the best of such work.

The key finding of the analysis is beautifully captured by this set of scatter plots:

Each dot is a state. The cases (left plot) and hospitalizations (right plot) are plotted against the severity of containment measures for November. The negative correlation is unmistakable: the more containment measures taken, the lower the counts.

There are a few features worth noting.

The severity index came from a group at Oxford, and is a number between 0 and 100. The journalists decided to leave out the numerical labels, instead simply showing More and Fewer. This significantly reduces processing time. Readers won't be able to understand the index values anyway without reading the manual.

The index values are doubly encoded. They are first encoded by the location on the horizontal axis and redundantly encoded on the blue-red scale. Ordinarily, I do not like redundant encoding because the reader might assume a third dimension exists. In this case, I had no trouble with it.

The easiest way to see the effect is to ignore the muddy middle and focus on the two ends of the severity index. Those states with the fewest measures - South Dakota, North Dakota, Iowa - are the worst in cases and hospitalizations while those states with the most measures - New York, Hawaii - are among the best. This comparison is similar to what is frequently done in scientific studies, e.g. when they say coffee is good for you, they typically compare heavy drinkers (4 or more cups a day) with non-drinkers, ignoring the moderate and light drinkers.

Notably, there is quite a bit of variability for any level of containment measures - roughly 50 cases per 100,000, and 25 hospitalizations per 100,000. This indicates that containment measures are not sufficient to explain the counts. For example, the hospitalization statistic is affected by the stock of hospital beds, which I assume differ by state.

Whenever we use a scatter plot, we run the risk of xyopia. This chart form invites readers to explain an outcome (y-axis values) using one explanatory variable (on x-axis). There is an assumption that all other variables are unimportant, which is usually false.

***

Because of the variability, the horizontal scale has meaningless precision. The next chart cures this by grouping the states into three categories: low, medium and high level of measures.

This set of charts extends the time window back to March 1. For the designer, this creates a tricky problem - because states adapt their policies over time. As indicated in the subtitle, the grouping is based on the average severity index since March, rather than just November, as in the scatter plots above.

***

The interplay between policy and health indicators is captured by connected scatter plots, of which the Times article included a few examples. Here is what happened in New York:

Up until April, the policies were catching up with the cases. The policies tightened even after the case-per-capita started falling. Then, policies eased a little, and cases started to spike again.

The Note tells us that the containment severity index is time shifted to reflect a two-week lag in effect. So, the case count on May 1 is not paired with the containment severity index of May 1 but of April 15.

***

You can find the full article here.

 

 

 

Making better pie charts if you must

I saw this chart on an NYU marketing twitter account:

The graphical design is not easy on our eyes. It's just hard to read for various reasons.

The headline sounds like a subject line from an email.

The subheaders are long, and differ only by a single word.

Even if one prefers pie charts, they can be improved by following a few guidelines.

First, start the first sector at the 12-oclock direction. Like this:

The survey uses a 5-point scale from "Very Good" to "Very Bad". Instead of using five different colors, it's better to use two extreme colors and shading. Like this:

I also try hard to keep all text horizontal.

For those who prefers not to use pie charts, a side-by-side bar chart works well.

In my article for DataJournalism.com, I outlined "unspoken rules" for making various charts, including pie charts.

 

 

 

Why you should expunge the defaults from Excel or (insert your favorite graphing program)

Yesterday, I posted the following chart in the post about Cornell's Covid-19 case rate after re-opening for in-person instruction.

This is an edited version of the chart used in Peter Frazier's presentation.

The original chart carries with it the burden of Excel defaults.

What did I change and why?

I switched away from the default color scheme, which ignores the relationships between the two lines. In particular, the key comparison on this chart should be the actual case rate versus the nominal case rate. In addition, the three lines at the top are related as they all come from the same underlying mathematical model. I used the same color but different shades.

Also, instead of placing the legend as far away from the data labels as possible, I moved the line labels next to the data labels.

Instead of daily date labels, I moved to weekly labels, and set the month names on a separate level than the day names.

The dots were removed from the top three lines but I'd have retained them, perhaps with some level of transparency, if I spent more time making the edits. I'd definitely keep the last dot to make it clear that the blue lines contain one extra dot.

***

Every graphing program has defaults, typically computed by some algorithm tuned to the average chart. Don't settle for the average chart. Get rid of any default setting that slows down understanding.

 

 

Election visual 3: a strange, mash-up visualization

Continuing our review of FiveThirtyEight's election forecasting model visualization (link), I now look at their headline data visualization. (The previous posts in this series are here, and here.)

It's a set of 22 maps, each showing one election scenario, with one candidate winning. What chart form is this?

Small multiples may come to mind. A small-multiples chart is a grid in which every component graphic has the same form - same chart type, same color scheme, same scale, etc. The only variation from graphic to graphic is the data. The data are typically varied along a dimension of interest, for example, age groups, geographic regions, years. The following small-multiples chart, which I praised in the past (link), shows liquor consumption across the world.

Each component graphic changes according to the data specific to a country. When we scan across the grid, we draw conclusions about country-to-country variations. As with convention, there are as many graphics as there are countries in the dataset. Sometimes, the designer includes only countries that are directly relevant to the chart's topic.

***

What is the variable FiveThirtyEight chose to vary from map to map? It's the scenario used in the election forecasting model.

This choice is unconventional. The 22 scenarios is a subset of the 40,000 scenarios from the simulation - we are left wondering how those 22 are chosen.

Returning to our question: what chart form is this?

Perhaps you're reminded of the dot plot from the previous post. On that dot plot, the designer summarized the results of 40,000 scenarios using 100 dots. Since Biden is the winner in 75 percent of all scenarios, the dot plot shows 75 blue dots (and 25 red).

The map is the new dot. The 75 blue dots become 16 blue maps (rounded down) while the 25 red dots become 6 red maps.

Is it a pictogram of maps? If we ignore the details on the maps, and focus on the counts of colors, then yes. It's just a bit challenging because of the hole in the middle, and the atypical number of maps.

As with the dot plot, the map details are a nice touch. It connects readers with the simulation model which can feel very abstract.

Oddly, if you're someone familiar with probabilities, this presentation is quite confusing.

With 40,000 scenarios reduced to 22 maps, each map should represent 1818 scenarios. On the dot plot, each dot should represent 400 scenarios. This follows the rule for creating pictograms. Each object in a pictogram - dot, map, figurine, etc. - should encode an equal amount of the data. For the 538 visualization, is it true that each of the six red maps represents 1818 scenarios? This may be the case but not likely.

Recall the dot plot where the most extreme red dot shows a scenario in which Trump wins 376 out of 538 electoral votes (margin = 214). Each dot should represent 400 scenarios. The visualization implies that there are 400 scenarios similar to the one on display. For the grid of maps, the following red map from the top left corner should, in theory, represent 1,818 similar scenarios. Could be, but I'm not sure.

Mathematically, each of the depicted scenario, including the blowout win above, occurs with 1/40,000 chance in the simulation. However, one expects few scenarios that look like the extreme scenario, and ample scenarios that look like the median scenario.  

So, the right way to read the 538 chart is to ignore the map details when reading the embedded pictogram, and then look at the small multiples of detailed maps bearing in mind that extreme scenarios are unique while median scenarios have many lookalikes.

(Come to think about it, the analogous situation in the liquor consumption chart is the relative population size of different countries. When comparing country to country, we tend to forget that the data apply to large numbers of people in populous countries, and small numbers in tiny countries.)

***

There's a small improvement that can be made to the detailed maps. As I compare one map to the next, I'm trying to pick out which states that have changed to change the vote margin. Conceptually, the number of states painted red should decrease as the winning margin decreases, and the states that shift colors should be the toss-up states.

So I'd draw the solid Republican (Democratic) states with a lighter shade, forming an easily identifiable bloc on all maps, while the toss-up states are shown with a heavier shade.

Here, I just added a darker shade to the states that disappear from the first red map to the second.