When words speak louder than pictures

I've been staring at this chart from the Wall Street Journal (link) about U.S. workers working remotely:

It's one of those offerings I think on which the designer spent a lot of effort, but ultimately didn't realize that the reader would spend equal if not more effort deciphering.

However, the following paragraph lifted straight from the article says exactly what needs to be said:

Workers overall spent an average of 5 hours and 25 minutes a day working from home in 2022. That is about two hours more than in 2019, the year before Covid-19 sent millions of workers scrambling to set up home oces, and down just 12 minutes from 2021, according to the Labor Department’s American Time Use Survey.

***

Why is the chart so hard to read?

It's mostly because the visual is fighting the message. In the Trifecta Checkup (link), this is represented by a disconnect between the Q(uestion) and the V(isual) corners - note the green arrow between these two corners.

The message concentrates on two comparisons: first, the increase in amount of remote work after the pandemic; and second, the mild decrease in 2022 relative to 2021.

On the chart, the elements that grab my attention are (a) the green and orange columns (b) the shading in the bottom part of those green and orange columns (c) the thick black line that runs across the chart (d) the indication on the left side that tells me one unit is an hour.

None of those visual elements directly addresses the comparisons. The first comparison - before and after the pandemic - is found by how much the green column spikes above the thick black line. Our comprehension is retarded by the decision to forego the typical axis labels in favor of chopping columns into one-hour blocks.

The second comparison - between 2022 and 2021 - is found in the white space above the top of the orange column.

So, in reality, the text labels that say exactly what needs to be said are carrying a lot of weight. A slight edit to the pointers helps connect those descriptions to the visual depiction, like this:

I've essentially flipped the tactics used in the various pointers. For the average level of remote work pre-pandemic, I dispense of any pointers while I'm using double-headed arrows to indicate differences across time.

Nevertheless, this modified chart is still too complex.

***

Here is a version that aligns the visual to the message:

It's a bit awkward because the 2 hour 48 minutes calculation is the 2021 number minus the average of 2015-19, skipping the 2020 year.

 

Why some dataviz fail

Maxim Lisnic's recent post should delight my readers (link). Thanks Alek for the tip. Maxim argues that charts "deceive" not merely by using visual tricks but by a variety of other non-visual means.

This is also the reasoning behind my Trifecta Checkup framework which looks at a data visualization project holistically. There are lots of charts that are well designed and constructed but fail for other reasons. So I am in agreement with Maxim.

He analyzed "10,000 Twitter posts with data visualizations about COVID-19", and found that 84% are "misleading" while only 11% of the 84% "violate common design guidelines". I presume he created some kind of computer program to evaluate these 10,000 charts, and he compiled some fixed set of guidelines that are regarded as "common" practice.

***

Let's review Maxim's examples in the context of the Trifecta Checkup.

The first chart shows Covid cases in the U.S. in July and August of 2021 (presumably the time when the chart was published) compared to a year ago (prior to the vaccination campaign).

Maxim calls this cherry-picking. He's right - and this is a pet peeve of mine, even with all the peer-reviewed scientific research. In my paper on problems with observational studies (link), my coauthors and I call for a new way forward: researchers should put their model calculations up on a website which is updated as new data arrive, so that we can be sure that the conclusions they published apply generally to all periods of time, not just the time window chosen for the publication.

Looking at the pair of line charts, readers can quickly discover its purpose, so it does well on the Q(uestion) corner of the Trifecta. The cherry-picking relates to the link between the Question and the Data, showing that this chart suffers from subpar analysis.

In addition, I find that the chart also misleads visually - the two vertical scales are completely different: the scale on the left chart spans about 60,000 cases while on the right, it's double the amount.

Thus, I'd call this a Type DV chart, offering opportunities to improve in two of the three corners.

***

The second chart cited by Maxim plots a time series of all-cause mortality rates (per 100,000 people) from 1999 to 2020 as columns.

The designer does a good job drawing our attention to one part of the data - that the average increase in all-cause mortality rate in 2020 over the previous five years was 15%. I also like the use of a different color for the pandemic year.

Then, the designer lost the plot. Instead of drawing a conclusion based on the highlighted part of the data, s/he pushed a story that the 2020 rate was about the same as the 2003 rate. If that was the main message, then instead of computing a 15% increase relative to the past five years, s/he should have shown how the 2003 and 2020 levels are the same!

On a closer look, there is a dashed teal line on the chart but the red line and text completely dominate our attention.

This chart is also Type DV. The intention of the designer is clear: the question is to put the jump in all-cause mortality rate in a historical context. The problem lies again with subpar analysis. In fact, if we take the two insights from the data, they both show how serious a problem Covid was at the time.

When the rate returned to the level of 2003, we have effectively gave up all the gains made over 17 years in a few months.

Besides, a jump in 15% from year to year is highly significant if we look at all other year-to-year changes shown on the chart.

***

The next section concerns a common misuse of charts to suggest causality when the data could only indicate correlation (and where the causal interpretation appears to be dubious). I may write a separate post about this vast topic in the future. Today, I just want to point out that this problem is acute with any Covid-19 research, including official ones.

***

I find the fourth section of Maxim's post to be less convincing. In the following example, the tweet includes two charts, one showing proportion of people vaccinated, and the other showing the case rate, in Iceland and Nigeria.

This data visualization is poor even on the V(isual) corner. The first chart includes lots of countries that are irrelevant to the comparison. It includes the unnecessary detail of fully versus partially vaccinated, unnecessary because the two countries selected are at two ends of the scale. The color coding is off sync between the two charts.

Maxim's critique is:

The user fails to account, however, for the fact that Iceland had a much higher testing rate—roughly 200 times as high at the time of posting—making it unreasonable to compare the two countries.

And the section is titled "Issues with Data Validity". It's really not that simple.

First, while the differential testing rate is one factor that should be considered, this factor alone does not account for the entire gap. Second, this issue alone does not disqualify the data. Third, if testing rate differences should be used to invalidate this set of data, then all of the analyses put out by official sources lauding the success of vaccination should also be thrown out since there are vast differences in testing rates across all countries (and also across different time periods for the same country).

One typical workaround for differential testing rate is to look at deaths rather than cases. For the period of time plotted on the case curve, Nigeria's cumulative death per million is about 1/8th that of Iceland. The real problem is again in the Data analysis, and it is about how to interpret this data casually.

This example is yet another Type DV chart. I'd classify it under problems with "Casual Inference". "Data Validity" is definitely a real concern; I just don't find this example convincing.

***

The next section, titled "Failure to account for statistical nuance," is a strange one. The example is a chart that the CDC puts out showing the emergence of cases in a specific county, with cases classified by vaccination status. The chart shows that the vast majority of cases were found in people who were fully vaccinated. The person who tweeted concluded that vaccinated people are the "superspreaders". Maxim's objection to this interpretation is that most cases are in the fully vaccinated because most people are fully vaccinated.

I don't think it's right to criticize the original tweeter in this case. If by superspreader, we mean people who are infected and out there spreading the virus to others through contacts, then what the data say is exactly that most such people are fully vaccinated. In fact, one should be very surprised if the opposite were true.

Indeed, this insight has major public health implications. If the vaccine is indeed 90% effective at stopping cases, we should not be seeing that level of cases. And if the vaccine is only moderately effective, then we may not be able to achieve "herd immunity" status, as was the plan originally.

I'd be reluctant to complain about this specific data visualization. It seems that the data allow different interpretations - some of which are contradictory but all of which are needed to draw a measured conclusion.

***
The last section on "misrepresentation of scientific results" could use a better example. I certainly agree with the message: that people have confirmation bias. I have been calling this "story-first thinking": people with a set story visualize only the data that support their preconception.

However, the example given is not that. The example shows a tweet that contains a chart from a scientific paper that apparently concludes that hydroxychloroquine helps treat Covid-19. Maxim adds this study was subsequently retracted. If the tweet was sent prior to the retraction, then I don't think we can grumble about someone citing a peer reviewed study published in Lancet.

***

Overall, I like Maxim's message. In some cases, I think there are better examples.

 

 

Visual story-telling: do you know or do you think?

One of the most important data questions of all time is: do you know? or do you think?

And one of the easiest traps to fall into is: I think, therefore I know.

***

Visual story-telling can be great but it can also mislead. Deception sometimes happens when readers are nudged to "fill in the blanks" with stuff they think they know, but they don't.

A Twitter reader asked me to look at the map in this Los Angeles Times (paywall) opinion column.

The column promptly announces its premise:

Years of widening economic inequality, compounded by the pandemic and political storm and stress, have given Americans the impression that the country is on the wrong track. Now there’s empirical data to show just how far the country has run off the rails: Life expectancies have been falling.

The writer creates the expectation that he will reveal evidence in the form of data to show that life expectancies have been driven down by economic inequality, pandemic, and politics. Does he succeed?

***

The map portrays average life expectancy (at birth) for some mysterious, presumably very recent, year for every county in the United States. From the color legend, we learn that the bottom-to-top range is about 20 years. There is a clear spatial pattern, with the worst results in the south (excepting south Florida).

The choice of colors is telling. Red and blue on a U.S. map has heavy baggage, as they signify the two main political parties in the country. Given that the author believes politics to be a key driver of health outcomes, the usage of red and blue here is deliberate. Throughout the article, the columnist connects the lower life expectancies in southern states to its politics.

For example, he said "these geographical disparities aren't artifacts of pure geography or demographics; they're the consequences of policy decisions at the state level... Of the 20 states with the worst life expectancies, eight are among the 12 that have not implemented Medicaid expansion under the Affordable Care Act..."

Casual readers may fall into a trap here. There is nothing on the map itself that draws the connection between politics and life expectancies; the idea is evoked purely through the red-blue color scheme. So, as readers, we are filling in the blanks with our own politics.

What could have been done instead? Let's look at the life expectancy map side by side with the map of the U.S. 2020 Presidential election.

Because of how close recent elections have been, we may think the political map has a nice balance of red and blue but it isn't. The Democrats' votes are heavily concentrated in densely-populated cities so most of the Presidential election map is red. When placed next to each other, it's obvious that politics don't explain the variance in life expectancy well. The Midwest is deep red and yet they have above average life expectancies. I have circled out various regions that contradict the claim that Republican politics drove life expectancies down.

It's not sufficient to point to the South, in which Republican votes and life expectancy are indeed inversely correlated. A good theory has to explain most of the country.

***

The columnist also suggests that poverty is the cause of low life expectancy. That too cannot be gleaned from the published map. Again, readers are nudged to use their wild imagination to fill in the blank.

Data come to the rescue. Here is a side-by-side comparison of the map of life expectancies and the map of median incomes.

A similar conundrum. While the story feels right in the South, it fails to explain the northwest, Florida, and various other parts of the country. Take a look again at the circled areas. Lower income brackets are also sometimes associated with high life expectancies.

***

The author supplies a third cause of lower life expectancies: Covid-19 response. Because Covid-19 was the "most obvious and convenient" explanation for the loss of life expectancy during the pandemic, this theory suggests that the red areas on the life expectancy map should correspond to the regions most ravaged by Covid-19.

Let's see the data.

The map on the right shows the number of confirmed cases until June 2021. As before, the correlation holds somewhat in the South but there are notable exceptions, e.g. the Midwest. We also have states with low Covid-19 cases but below-average life expectancy.

***

What caused the decline of life expectancy in the U.S. - which began before the pandemic, and has continued beyond - is highly complex, beyond what a single map or a pair of maps or a few pairs of maps could convey. Showing a red-blue map presents a trap for readers to fall into, in which they start thinking, without knowing.

 

Trying too hard

Today, I return to the life expectancy graphic that Antonio submitted. In a previous post, I looked at the bumps chart. The centerpiece of that graphic is the following complicated bar chart.

Let's start with the dual axes. On the left, age, and on the right, year of birth. I actually like this type of dual axes. The two axes present two versions of the same scale so the dual axes exist without distortion. It just allows the reader to pick which scale they want to use.

It baffles me that the range of each bar runs from 2.5 years to 7.5 years or 7.5 years to 2.5 years, with 5 or 10 years situated in the middle of each bar.

Reading the rest of the chart is like unentangling some balled up wires. The author has created a statistical model that attributes cause of death to male life expectancy in such a way that you can take the difference in life expectancy between two time points, and do a kind of waterfall analysis in which each cause of death either adds to or subtracts from the prior life expectancy, with the sum of these additions and substractions leading to the end-of-period life expectancy.

The model is complicated enough, and the chart doesn't make it any easier.

The bars are rooted at the zero value. The horizontal axis plots addition or substraction to life expectancy, thus zero represents no change during the period. Zero does not mean the cause of death (e.g. cancer) does not contribute to life expectancy; it just means the contribution remains the same.

The changes to life expectancy are shown in units of months. I'd prefer to see units of years because life expectancy is almost always given in years. Using years turn 2.5 months into 0.2 years which is a fraction, but it allows me to see the impact on the reported life expectancy without having to do a month-to-year conversion.

The chart highlights seven causes of death with seven different colors, plus gray for others.

What really does a number on readers is the shading, which adds another layer on top of the hues. Each color comes in one of two shading, referencing two periods of time. The unshaded bar segments concern changes between 2010 and "2019" while the shaded segments concern changes between "2019" and 2020. The two periods are chosen to highlight the impact of COVID-19 (the red-orange color), which did not exist before "2019".

Let's zoom in on one of the rows of data - the 72.5 to 77.5 age group.

COVID-19 (red-orange) has a negative impact on life expectancy and that's the easy one to see. That's because COVID-19's contribution as a cause of death is exactly zero prior to "2019". Thus, the change in life expectancy is a change from zero. This is not how we can interpret any of the other colors.

Next, we look at cancer (blue). Since this bar segment sits on the right side of zero, cancer has contributed positively to change in life expectancy between 2010 and 2020. Practically, that means proportionally fewer people have died from cancer. Since the lengths of these bar segments correspond to the relative value, not absolute value, of life expectancy, longer bars do not necessarily indicate more numerous deaths.

Now the blue segment is actually divided into two parts, the shaded and not shaded. The not-shaded part is for the period "2019" to 2020 in the first year of the COVID-19 pandemic. The shaded part is for the period 2010 to "2019". It is a much wider span but it also contains 9 years of changes versus "1 year" so it's hard to tell if the single-year change is significantly different from the average single-year change of the past 9 years. (I'm using these quotes because I don't know whether they split the year 2019 in the middle since COVID-19 didn't show up till the end of that year.)

Next, we look at the yellow-brown color correponding to CVD. The key feature is that this block is split into two parts, one positive, one negative. Prior to "2019", CVD has been contributing positively to life expectancy changes while after "2019", it has contributed negatively. This observation raises some questions: why would CVD behave differently with the arrival of the pandemic? Are there data problems?

***

A small multiples design - splitting the period into two charts - may help here. To make those two charts comparable, I'd suggest annualizing the data so that the 9-year numbers represent the average annual values instead of the cumulative values.

 

 

Two uses of bumps charts

Long-time reader Antonio R. submitted the following chart, which illustrates analysis from a preprint on the effect of Covid-19 on life expectancy in the U.S. (link)

For this post, I want to discuss the bumps chart on the lower right corner. Bumps charts are great at showing change over time. In this case, the authors are comparing two periods "2010-2019" and "2019-2020". By glancing at the chart, one quickly divides the causes of death into three groups: (a) COVID-19 and CVD, which experienced a big decline (b) respiratory, accidents, others ("rest"), and despair, which experienced increases, and (c) cancer and infectious, which remained the same.

And yet, something doesn't seem right.

What isn't clear is the measured quantity. The chart title says "months gained or lost" but it takes a moment to realize the plotted data are not number of months but ranks of the effects of the causes of deaths on life expectancy.

Observe that the distance between each cause of death is the same. Look at the first rising line (respiratory): the actual values went from 0.8 months down to 0.2.

***

While the canonical bumps chart plots ranks, the same chart form can be used to show numeric data. I prefer to use the same term for both charts. In recent years, the bumps chart showing numeric data has been called "slopegraph".

Here is a side-by-side comparison of the two charts:

The one on the left is the same as the original. The one on the right plots the number of months increased or decreased.

The choice of chart form paints very different pictures. There are four blue lines on the left, indicating a relative increase in life expectancy - these causes of death contributed more to life expectancy between the two periods. Three of the four are red lines on the right chart. Cancer was shown as a flat line on the left - because it was the highest ranked item in both periods. The right chart shows that the numeric value for cancer suffered one of the largest drops.

The left chart exaggerates small numeric changes while it condenses large numeric changes.

 

 

Another reminder that aggregate trends hide information

The last time I looked at the U.S. employment situation, it was during the pandemic. The data revealed the deep flaws of the so-called "not in labor force" classification. This classification is used to dehumanize unemployed people who are declared "not in labor force," in which case they are neither employed nor unemployed -- just not counted at all in the official unemployment (or employment) statistics.

The reason given for such a designation was that some people just have no interest in working, or even looking for a job. Now they are not merely discouraged - as there is a category of those people. In theory, these people haven't been looking for a job for so long that they are no longer visible to the bean counters at the Bureau of Labor Statistics.

What happened when the pandemic precipitated a shutdown in many major cities across America? The number of "not in labor force" shot up instantly, literally within a few weeks. That makes a mockery of the reason for such a designation. See this post for more.

***

The data we saw last time was up to April, 2020. That's more than two years old.

So I have updated the charts to show what has happened in the last couple of years.

Here is the overall picture.

In this new version, I centered the chart at the 1990 data. The chart features two key drivers of the headline unemployment rate - the proportion of people designated "invisible", and the proportion of those who are considered "employed" who are "part-time" workers.

The last two recessions have caused structural changes to the labor market. From 1990 to late 2000s, which included the dot-com bust, these two metrics circulated within a small area of the chart. The Great Recession of late 2000s led to a huge jump in the proportion called "invisible". It also pushed the proportion of part-timers to all0time highs. The proportion of part-timers has fallen although it is hard to interpret from this chart alone - because if the newly invisible were previously part-time employed, then the same cause can be responsible for either trend.

Readers of Numbersense (link) might be reminded of a trick used by school deans to pump up their US News rankings. Some schools accept lots of transfer students. This subpopulation is invisible to the US News statisticians since they do not factor into the rankings. The recent scandal at Columbia University also involves reclassifying students (see this post).

Zooming in on the last two years. It appears that the pandemic-related unemployment situation has reversed.

***

Let's split the data by gender.

American men have been stuck in a negative spiral since the 1990s. With each recession, a higher proportion of men are designated BLS invisibles.

In the grid system set up in this scatter plot, the top right corner is the worse of all worlds - the work force has shrunken and there are more part-timers among those counted as employed. The U.S. men are not exiting this quadrant any time soon.

***
What about the women?

If we compare 1990 with 2022, the story is not bad. The female work force is gradually reaching the same scale as in 1990 while the proportion of part-time workers have declined.

However, celebrating the above is to ignore the tremendous gains American women made in the 1990s and 2000s. In 1990, only 58% of women are considered part of the work force - the other 42% are not working but they are not counted as unemployed. By 2000, the female work force has expanded to include about 60% with similar proportions counted as part-time employed as in 1990. That's great news.

The Great Recession of the late 2000s changed that picture. Just like men, many women became invisible to BLS. The invisible proportion reached 44% in 2015 and have not returned to anywhere near the 2000 level. Fewer women are counted as part-time employed; as I said above, it's hard to tell whether this is because the women exiting the work force previously worked part-time.

***

The color of the dots in all charts are determined by the headline unemployment number. Blue represents low unemployment. During the 1990-2022 period, there are three moments in which unemployment is reported as 4 percent or lower. These charts are intended to show that an aggregate statistic hides a lot of information. The three times at which unemployment rate reached historic lows represent three very different situations, if one were to consider the sizes of the work force and the number of part-time workers.

 

P.S. [8-15-2022] Some more background about the visualization can be found in prior posts on the blog: here is the introduction, and here's one that breaks it down by race. Chapter 6 of Numbersense (link) gets into the details of how unemployment rate is computed, and the implications of the choices BLS made.

P.S. [8-16-2022] Corrected the axis title on the charts (see comment below). Also, added source of data label.

Think twice before you spiral

After Nathan at FlowingData sang praises of the following chart, a debate ensued on Twitter as others dislike it.

The chart was printed in an opinion column in the New York Times (link).

I have found few uses for spiral charts, and this example has not changed my mind.

The canonical time-series chart is like this:

 

***

The area chart takes no effort to understand. We can see when the peaks occurred. We notice that the current surge is already double the last peak seen a year ago.

It's instructive to trace how one gets from the simple area chart to the spiral chart.

Step 1 is to center the area on the zero baseline, instead of having the zero baseline as the baseline. While this technique frequently makes for a more pleasant visual (because of our preference for symmetry), it actually makes it harder to see the trend over time. Effectively, any change is split in half, which is why the envelope of the area is less sharp.

In Step 2, I massively compress the vertical scale. That's because when you plot a spiral, you are forced to fit each cycle of data into a much shorter range. Such compression causes the year on year doubling of cases to appear less dramatic. (Actually, the aspect ratio is devastated because while the vertical scale is hugely compressed, the horizontal scale is dramatically stretched out due to the curled up design)

Step 3 may elude your attention. If you simply curl up the compressed, centered area chart, you don't get the spiral chart. The key is to ask about the radius of the spiral. As best I can tell, the radius has no meaning; it is gradually increased so that each year of data has its own "orbit". What would the change in radius translate to on our non-circular chart? It should mean that the center of the area is gradually lifted away from the zero line. On the right chart, I mimic this effect (I only measured the change in radius every 3 months so the change is more angular than displayed in the spiral chart.) The problem I have with this Step is that it serves no purpose, while it complicates cognition,

In Step 4, just curl up the object into a ball based on aligning months of the year.

This is the point when I realized I missed a Step 2B. I carefully aligned the scales of both charts so that the 150K cases shown in the legend on the right have the same vertical representation as on the left. This exposes a severe horizontal rescaling. The length of the horizontal axis on the left chart is many times smaller than the circumference of the spiral! That's why earlier, I said one of the biggest feature of this spiral chart is that it imposes a dubious aspect ratio, that is extremely wide and extremely short.

As usual, think twice before you spiral.

 

 

Illustrating coronavirus waves with moving images

The New York Times put out a master class in visualizing space and time data recently, in a visualization of five waves of Covid-19 that have torched the U.S. thus far (link).

The project displays one dataset using three designs, which provides an opportunity to compare and contrast them.

***

The first design - above the headline - is an animated choropleth map. This is a straightforward presentation of space and time data. The level of cases in each county is indicated by color, dividing the country into 12 levels (plus unknown). Time is run forward. The time legend plays double duty as a line chart that shows the change in the weekly rate of reported cases over the course of the pandemic. A small piece of interactivity binds the legend with the map.

(To see a screen recording of the animation, click on the image above.)

***

The second design comprises six panels, snapshots that capture crucial "turning points" during the Covid-19 pandemic. The color of each county now encodes an average case rate (I hope they didn't just average the daily rates). 

The line-chart legend is gone -  it's not hard to see Winter > Fall 2020 > Summer/Fall 2021 >... so I don't think it's a big loss.

The small-multiples setup is particularly effective at facilitating comparisons: across time, and across space. It presents a story in pictures.

They may have left off 2020 following "Winter" because December to February spans both years but "Winter 2020" may do more benefit than harm here.

***

The third design is a series of short films, which stands mid-way between the single animated map and the six snapshots. Each movie covers a separate window of time.

This design does a better job telling the story within each time window while it obstructs comparisons across time windows.

The informative legend is back. This time, it's showing the static time window for each map.

***

The three designs come from the same dataset. I think of them as one long movie, six snapshots, and five short films.

The one long movie is a like a data dump. It shows every number in the dataset, which is the weekly case rate for each county for a given week. All the data are streamed into a single map. It's a show piece.

As an instrument to help readers understand the patterns in the dataset, the movie falls short. Too much is going on, making it hard to focus and pick out key trends. When your eyes are everywhere, they are nowhere.

The six snapshots represent the other extreme. The graph does not move, as the time axis is reduced to six discrete time points. But this display describes the change points, and tells a story. The long movie, by contrast, invites readers to find a story.

Without motion, the small-multiples format allows us to pick out specific counties or regions and compare the case rates across time. This task is close to impossible in the long movie, as it requires freezing the movie, and jumping back and forth.

The five short films may be the best of both worlds. It retains the motion. If the time windows are chosen wisely, each short film contains a few simple patterns that can easily be discerned. For example, the third film shows how the winter wave emerged from the midwest and then walloped the whole country, spreading southward and toward the coasts.

(If the above gif doesn't play, click it.)

***

If there is double or triple the time allocated to this project, I'd want to explore spatial clustering. I'd like to dampen the spatial noise (neighboring counties that have slightly different experiences). There is also temporal noise (fluctuations from week to week for the same county) - which can be smoothed away. I think with these statistical techniques, the "wave" feature of the pandemic may be more visible.

 

 

Visually displaying multipliers

As I'm preparing a blog about another real-world study of Covid-19 vaccines, I came across the following chart (the chart title is mine).

As background, this is the trend in Covid-19 cases in the U.K. in the last couple of months, courtesy of OurWorldinData.org.

The React-1 Study sends swab kits to randomly selected people in England in order to assess the prevalence of Covid-19. Every month, there is a new round of returned swabs that are tested for Covid-19. This measurement method captures asymptomatic cases although it probably missed severe and hospitalized cases. Despite having some shortcomings, this is a far better way to measure cases than the hotch-potch assembling of variable-quality data submitted by different jurisdictions that has become the dominant source of our data.

Rounds 12 and 13 captured an inflection point in the pandemic in England. The period marked the beginning of the end of the belief that widespread vaccination will end the pandemic.

The chart I excerpted up top broke the data down by age groups. The column heights represent the estimated prevalence of Covid-19 during each round - also, described precisely in the paper as "swab positivity." Based on the study's design, one may generalize the prevalence to the population at large. About 1.5% of those aged 13-24 in England are estimated to have Covid-19 around the time of Round 13 (roughly early July).

The researchers came to the following conclusion:

We show that the third wave of infections in England was being driven primarily by the Delta variant in younger, unvaccinated people. This focus of infection offers considerable scope for interventions to reduce transmission among younger people, with knock-on benefits across the entire population... In our data, the highest prevalence of infection was among 12 to 24 year olds, raising the prospect that vaccinating more of this group by extending the UK programme to those aged 12 to 17 years could substantially reduce transmission potential in the autumn when levels of social mixing increase

***

Raise your hand if the graphics software you prefer dictates at least one default behavior you can't stand. I'm sure most hands are up in the air. No matter how much you love the software, there is always something the developer likes that you don't.

The first thing I did with today's chart is to get rid of all such default details.

For me, the bottom chart is cleaner and more inviting.

***

The researchers wanted readers to think in terms of Round 3 numbers as multiples of Round 2 numbers. In the text, they use statements such as:

weighted prevalence in round 13 was nine-fold higher in 13-17 year olds at 1.56% (1.25%, 1.95%) compared with 0.16% (0.08%, 0.31%) in round 12

It's not easy to perceive a nine-fold jump from the paired column chart, even though this chart form is better than several others. I added some subtle divisions inside each orange column in order to facilitate this task:

I have recommended this before. I'm co-opting pictograms in constructing the column chart.

An alternative is to plot everything on an index scale although one would have to drop the prevalence numbers.

***

The chart requires an additional piece of context to interpret properly. I added each age group's share of the population below the chart - just to illustrate this point, not to recommend it as a best practice.

The researchers concluded that their data supported vaccinating 13-17 year olds because that group experienced the highest multiplier from Round 12 to Round 13. Notice that the 13-17 year old age group represents only 6 percent of England's population, and is the least populous age group shown on the chart.

The neighboring 18-24 age group experienced a 4.5 times jump in prevalence in Round 13 so this age group is doing much better than 13-17 year olds, right? Not really.

While the same infection rate was found in both age groups during this period, the slightly older age group accounted for 50% more cases -- and that's due to the larger share of population.

A similar calculation shows that while the infection rate of people under 24 is about 3 times higher than that of those 25 and over, both age groups suffered over 175,000 infections during the Round 3 time period (the difference between groups was < 4,000).  So I don't agree that focusing on 13-17 year olds gives England the biggest bang for the buck: while they are the most likely to get infected, their cases account for only 14% of all infections. Almost half of the infections are in people 25 and over.

 

Working hard at clarity

As I am preparing another blog post about the pandemic, I came across the following data graphic, recently produced by the CDC for a vaccine advisory board meeting:

This is not an example of effective visual communications.

***

For one thing, readers are directed to scour the footnotes to figure out what's going on. If we ignore those for the moment, we see clusters of bubbles that have remained pretty stable from December 2020 to August 2021. The data concern some measure of Americans' intent to take the COVID-19 vaccine. That much we know.

There may have been a bit of an upward trend between January and May, although if you were shown the clusters for December, February and April, you'd think the trend's been pretty flat. 

***

But those colors? What could they represent? You'd surely have to fish this one out of the footnotes. Specifically, this obtuse sentence: "Surveys with multiple time points are shown with the same color bubble for each time point." I had to read it several times. I think it simply means "Color represents the pollster." 

Then it adds: "Surveys with only one time point are shown in gray." which simply means "All pollsters who have only one entry in the dataset are grouped together and shown in gray."

Another problem with this chart is over-plotting. Look at the July cluster. It's impossible to tell how many polls were conducted in July because the circles pile on top of one another. 

***

The appearance of the flat trend is a result of two unfortunate decisions made by the designer. If I retained the chart form, I'd have produced something that looks like this:

The first design choice is to expand the vertical axis to range from 0% to 100%. This effectively squeezes all the bubbles into a small range.

The second design choice is to enlarge the bubbles causing copious amount of overlapping. 

In particular, this decision blows up the Pew poll (big pink bubble) that contained 10 times the sample size of most of the other polls. The Pew outcome actually came in at 70% but the top of the pink bubble extends to over 80%. Because of this, the outlier poll of December 2020 - which surprisingly printed the highest number of all polls in the entire time window - no longer looks special. 

***

Now, let's see what else we can do to enhance this chart. 

I don't like how bubble size is used to encode the sample size. It creates a weird sensation for anyone who's familiar with sampling errors, and confidence regions. The Pew poll with 10 times the sample size is the most reliable poll of them all. Reliability means the error bars around the Pew poll outcome is the smallest of them all. I tend to think of the area around a point estimate as showing the sampling error so the Pew poll would be a dot, showing the high precision of that estimate. 

But that won't work because larger bubbles catch more of the reader's attention. So, in the following version, all dots have the same size. I encode reliability in the opacity of the color. The darker dots are polls that are more reliable, that have larger sample sizes.

Two of the pollsters have more frequent polling than others. In this next version, I highlighted those two, which reveals the trend better.