line chart – Page 10 – Coffee Spoons

Flood Stages of the Schuylkill

Hurricane Isaias ran up the East Coast of the United States then the Hudson River Valley before entering Canada. Before it left the US, however, it dumped some record-setting amounts of rain in Philadelphia and across the region. And in times of heavy rains, the lower-lying areas of the city (and suburbs like Upper Darby and Downingtown to mention a few) face inundation from swollen rivers and creeks. And in the city itself, the neighbourhood of Eastwick is partially built upon a floodplain. So staying atop river levels is important and the National Weather Service has been doing that for years.

The National Weather Service graphic above is from this very morning and represents the water level of the Schuylkill River (the historical Philadelphia was sited between two rivers, the more commonly known Delaware and its tributary the Schuylkill), which receives water from the suburbs to the north and west of the city, the area hardest hit by Isaias’ rainfall.

The chart looks at the recent as well as the forecast stages of the river. Not surprisingly, the arrival of Isaias accounts for the sudden rise in the blue line. But there is a lot going on here, yellows, reds, and purples, some kind of NOAA logo behind the chart, labels sitting directly on lines, and some of the type is pixellated and difficult to read.

But it does do a nice job of showing the differences in observations and forecast points in time. By that I mean, a normal line chart has an equal distribution of observations along its length. There is an equal space between the weeks or the months or the years. But in instances like this, observations may not be continuous—imagine a flood destroying a sensor—or here that the forecasts are not as frequently produced as observations. And so these are all called out by the dots on the lines we see.

This is the chart I am accustomed to seeing. But then last night, reading about the damage I came across this graphic (screenshot also from this morning to compare to above) from the Philadelphia Inquirer.

It takes the same data and presents it a cleaner, clearer fashion. The flood stages are far easier to read. Gone is the NOAA logo and the unnecessary vertical gridlines. The type is far more legible and the palette less jarring and puts the data series in front and centre.

In general, this is a tremendous improvement for the legibility of the chart. I would probably use a different colour for the record flood stage line, or given their use of solid lines for the axis maybe make it dotted. But that’s a small quibble.

The only real issue here is what happens to the time? Compare the frequent observations in the past in the original, every half hour or so, to the six hourly dots (the blue versus the purple). In the Inquirer version, those spaces between forecast points disappear and become the same as the half-hour increments.

To be fair, the axis labelleing implies this as the label goes from August 4 to 5 and then jumps all the way to 7, but it is not as intuitive as it could be. Here I would recommend following the National Weather Service’s fashion of adjusting for the time gap. It would probably mean some kind of design tweak to emphasise that the observations earlier than now are observed every half hour or so, versus the six-hour forecasts. The NWS did this through dots. One could use a dotted line, or some other design treatment.

This missing time is the only thing really holding back this piece from the Inquirer from standing out as a great update of the traditional National Weather Service hydrograph chart.

Credit for the National Weather Service piece goes to the National Weather Service.

Credit for the Inquirer piece goes to Dominique DeMoe.

Big Bar Chart Better

Today isn’t a Friday, but I want to take a quick look at something that made me laugh aloud—literally LOL—whilst simultaneously cringe.

Not surprisingly it has to do with Trump and data/facts.

This all stems from an interview Axios’ Jonathan Swan conducted with President Trump on 28 July and that was released yesterday. I haven’t watched the interview in its entirety, but I’ve seen some excerpts. Including this gem.

It’s eerily reminiscent of a British show called The Thick of It written by Armando Iannucci or probably more accurately an interview out of one his earlier works with Chris Morris, On the Hour or The World Today. He later went on to create Veep for American audiences, based loosely or inspired by the Thick of It, but I found it a weak substitute for the original. But I digress.

In that clip, the President talks about how he looks at the number of deaths as a share of cases, the case fatality rate, whilst Swan is discussing deaths as a share of total population, deaths per capita. Now the latter is not a great data point to use, especially in the middle of the pandemic, because we’re not certain what the actual denominator is. I’ve discussed this before in some of my “this is not the flu” posts where the case fatality rate, sometimes more commonly called simply the mortality rate, was in the 3–5% range.

Regardless of whether or not one should use the metric, here is how the President visualised that data.

Four big and beautiful bar charts. The best charts.

The President claims the United States “Look, we’re last. Meaning we’re first. We have the best. Take a look again, it’s cases [it’s actually still the case mortality rate]. And we have cases because of the testing.”

The problem is that one, it’s the wrong metric. Two, the idea that testing creates cases is…insane. Third, the United States is last in that big set of bar charts. Why is every country a different colour? In the same data series, they should all be the same, unless you’re encoding a variable such as, say, region via colour. But with four data points, a bar chart taking up the entirety of a US-letter sized paper is grossly inefficient.

But that’s not even the full picture. Because if you look at a more robust data set, this one from Our World in Data, we get a better sense of where the United States sits.

Still not the highest on the chart, true. But even in this set; Norway (of not a shithole fame), India, South Korea, New Zealand, South Africa, and Congo all rank lower. The United States is far from last. And for those wondering, yes, I took the data from the same date as the interview.

There’s another clip within that clip I linked to earlier that deals with South Korea’s numbers and how the President says we “don’t know that”. And this is the bigger problem. We all know that data can be manipulated. But if we cannot agree that the data is real, we cannot have a framework for a real discourse on how to solve very real problems.

As someone who works with data to communicate information or stories on a near daily basis, this is just frightening. It’s as if you say to me, the sky is a beautiful shade of blue today without a cloud in the sky and I reply, no, I think it’s a foreboding sky with those heavy clouds of green with red polka dots. At that point we cannot even have a discussion about the weather.

And it’s only Tuesday.

Credit for the Trump graphic goes to somebody in the White House I assume.

Credit for the complete graphics goes to Our World in Data.

Covid-19 Update

As I mentioned last week, I am going to try using my blog here for the weekly update on the five states people have asked me to explore. And for the second week in a row, we are basically seeing numbers down compared to previous days. But given that numbers are generally lower on the weekends, that is not terribly surprising.

The real question is by Friday, will these numbers have rebounded?

The Covid-19 curves for PA, NJ, DE, VA, and IL

The Covid-19 death curves for PA, NJ, DE, VA, and IL.

Credit for these graphics is mine.

Sunday’s Covid Numbers

I do not want this blog to become a permanent Covid-19 data site. So in my push to resume posting last week, I tried to keep to from posting the numbers and instead focused on discussing how the data is displayed.

But I hear from quite a few people via comments, DMs, emails, and text messages that they find the graphics I produce helpful. So on the blog, I’m going to try posting just one set of graphics per week. Will it always be Monday? I don’t know. On the one hand, new week, new data. But on the other, weekend numbers tend to be lower than the rest of the week and could make it seem like, yay, the numbers are starting to go down especially if you only come to my blog and only see this data once a week.

Daily cases and their rolling average for Pennsylvania, New Jersey, Delaware, Virginia, and Illinois — Daily new cases

Daily new deaths and their rolling average for Pennsylvania, New Jersey, Delaware, Virginia, and Illinois. — Daily new deaths

So yeah, we’ll see how this goes. And I’ll try to keep Tuesday–Friday to discussing the world of data visualisation, although in these days, a good chunk of it will likely revolve around Covid.

Credit for these graphics is mine.

Corona Curves

It’s Friday. I’d normally say something like we’ve survived this far, but the fact of the matter is that thousands have not. But, still, let’s try to keep it a little light. So here’s something from xkcd about the shape of the various curve potentials for Covid-19.

PA seems to be somewhere around Scenario 2.

Credit for the piece goes to Randall Munroe.

Wednesday’s Covid-19 Data

Here we have the data from Wednesday for Covid-19.

Pennsylvania saw continued spread of the virus. Notably, Monroe County in eastern Pennsylvania passed 1000 cases. It was one of the state’s earliest hotspots. That appears to have been because it was advertised as a corona respite for people from New York, not too far to the east and by then in the grips of their own outbreak.

New Jersey grimly passed 5000 deaths Wednesday. And it is on track to pass 100,000 total cases likely Friday or Saturday. Almost 2/3 of these cases are located in North Jersey, with some South Jersey counties still reporting just a few hundred cases and a handful of deaths.

Delaware passed 3000 cases and Kent Co. passed 500. While those don’t read like large numbers, keep in mind the relatively small population of the state.

Virginia has restarted reporting deaths, this time at the county level and not the health district level. What we see is deaths being reported all over the eastern third of the state from DC through Richmond down to Virginia Beach. In the interior counties we are beginning to see the first deaths appear. And in western counties, we still see that the virus has yet to reach some locations, but counties are beginning to report their first cases.

Illinois continues to suffer greatly in the Chicago area, and at levels that dwarf the remainder of the state. However, the downstate counties are beginning to see spikes of their own. Macon and Jefferson Counties each saw increases of 30–40 cases in just 24 hours.

Preview(opens in a new tab)

A longer-term look at the states shows how the states diverge in their outbreaks. Pennsylvania looks like it might be forcing the curve downward whereas New Jersey appears to have more plateaued. Earlier I expressed concern about Virginia, which does now appear to have not peaked and continues to see an increasing rate of spread. Then we have Illinois, which may have plateaued, but we need to see if yesterday’s record amount of new cases was a blip or an inflection point. And in Delaware a missing day of records makes it tricker to see what exactly the trend is.

Credit for the piece is mine.

Covid-19 Data from Monday

Monday’s Covid-19 data for Pennsylvania, New Jersey, Delaware, Virginia, and Illinois provided a glimmer of good news, most notably in Pennsylvania. That, however, occurred on the same day as a protest in Harrisburg that could set the state back days if not weeks. More on that below.

Pennsylvania saw fewer than 1000 new cases for the first time since 1 April. The curve here may be doing more than flattening, but it might actually be falling. That is to say the infection rate is decreasing rather than stabilising and holding steady, as it appears to be doing in New Jersey. That said, new cases are appearing sporadically in the rural and less dense areas of the state. Problematically, protestors arrived in Harrisburg to let it be known they are unhappy with the quarantine. Because the rest of us are.

The problem is that it appears a significant percentage of those infected with the virus are asymptomatic carrier, i.e. they are sick, but do not show any symptoms like fever, coughing, difficulty breathing. Critically, they may not appear sick, but they can spread the sickness. And so a gathering of several hundred people in close quarters? Not ideal.

Compare that to a Christian cultish church in Daegu, South Korea. There, an infected parishioner did not heed government calls to isolate and instead attended a church service. The average infected person spreads this virus to two or three people. This congregant? They infected 43 people who then went on to infect other people.

It is quite possible that someone in that Harrisburg protest was an asymptomatic carrier. And given the lack of social distancing, the lack of masks, and the general reckless behaviour, it is quite possible that the rally could be a super-spreading event. But we won’t know for 5–10 days, the apparent incubation period of the virus. Hopefully we dodge the proverbial bullet. But it is quite easy to see how these kinds of protests could lead to surges in infections. And those surges would then force the government to extend its quarantine by weeks thereby defeating the entire point of the protestors.

We get it. Quarantine sucks. But we all have to suck it up.

Moving on to New Jersey, where we see continuing evidence of the plateauing of cases. The bulk of the cases remain in the north in the New York suburban counties with the fewest numbers in the counties in South Jersey. However, averages of nearly 3500 new cases daily remains quite high and the death toll of 4377 is likely to continue to climb higher, even if Monday’s 175 new deaths was lower than most days in recent weeks.

Delaware is back to reporting its figures. And in that release, we had Sussex County in the south climb above 1000 total cases. The levels or curves chart at the end will also show how the state might be flattening and stabilising its infection rate, but we will need several days of uninterrupted reporting to make that determination.

Virginia might be worrying. Or it might not be. Cases continue to increase in the big metropolitan counties like Fairfax and Henrico. But, there are still several counties out in the west that remain unaffected. And the curves chart at the end shows how there has not yet been any sort of even a near-exponential growth curve. Instead we just see a steady, slow increase in the number of cases. That in its own way makes it more difficult to see when the curve flattens, because it was already a relatively flat curve.

Illinois continues to be the tale of two states: Chicago vs. everywhere else. The combined Chicago and Cook County have over 20,000 total cases and the surrounding counties add a few thousand more, which gets you over 2/3 of the state’s 31,000 cases. That said, new cases and new fatalities are beginning to pop up in downstate counties.

Lastly a look at the curves. As I noted above when talking about Pennsylvania, you can clearly see the downward slope of the state’s new cases curve. Compare that to the plateau-like shape of New Jersey. Delaware and Illinois might be approaching a New Jersey-like curves. But I would want to see more data and in Delaware less volatility. But like I said, Virginia is a tricky one to read.

Credit for the pieces is mine.

Sunday’s Covid-19 Data

Here is a look at the data from Sunday’s releases on the COVID-19 outbreak in Pennsylvania, New Jersey, Virginia, and Illinois. I’ve omitted Delaware because they paused reporting on Sunday to move to a noontime release instead of their previous end-of-day.

I’m not exactly certain what that means for the data on Delaware and reporting time series. But, my guess is that will be more like a hole in the time series. I need to spend some time looking at that. But, anyways, on to the states for which we did have data on Sunday.

Pennsylvania continues to see growth in cases, but as we’ll get to with the levels, that appears to at least be stabilising. But in the spread of the outbreak, we are beginning to see the T of the state, that more rural and less densely populated area, beginning to fill in with cases. These are of course the areas of concern, the areas with shuttered rural hospitals, lack of comparatively developed infrastructure, where the impacts could be proportionately more severe than in the bigger cities. In terms of deaths, they have now spread almost across the state from east to west. I am still waiting until two adjacent counties connect before I make that final pronouncement.

For New Jersey, I have removed the orange outlines around each county. The initial idea was to show where deaths had occurred. But now that they have been reported in every county, they don’t seem to be as helpful as the small number I provide in the graphic. Regardless, 4200 deaths is a lot. But the approximately 200 new deaths is the lowest number reported in several days.

Virginia is a weird state. When we see the levels chart below, you will see how its uptick has been far more gradual, and to this point it does not yet appear to have peaked or begun to stabilise. Most of the reported cases continue to be in and around the state’s big cities, notably the DC metro area, but also Richmond.

Illinois has now seen cases from north at the Wisconsin border all the way south to Cairo. Most cases remain, however, concentrated in the Chicago metropolitan area, with lesser scale outbreaks occurring in the Quad Cities area and the suburban counties of Illinois this side of the Mississippi. Deaths continue to rise, and while most area again in the Chicago area, they are appearing increasingly at low levels in downstate counties.

But what about those curves? Excepting Delaware, which hasn’t reported new data, we can see that some states like Virginia continue to see increasing rates of infection. Others like New Jersey and Pennsylvania clearly have flattened and have entered a new phase. In New Jersey’s case it appears to be more of a stabilised plateau. In Pennsylvania, there was some evidence it was entering a declining rate phase, but that may now have begun to become more of a steady rate of infection like in New Jersey. Illinois is tough to read because of the variability of its data. It might be more of a pause in the rate of increase, or it may have begun to stabilise. We need more data.

Credit for the pieces is mine.

Monday Covid-19 Data

The data from Monday provided yet more evidence that the outbreak is flattening in several states. However, in some, the outbreak continues to pick up steam. Does this runs contrary to the idea that as a country is flattening? Not necessarily, but it is important to remember that a country that spans a continent and holds 330 million people will experience the pandemic differently at different times. So some states like Washington will be first, and others will be last.

Our five states cover the range of worsening to stabilising. We hope that those stabilising states soon enter the improving phase. Though to beat the dead horse, I would add that just because a state is improving doesn’t mean we can all go back to life like we knew it two months ago. That would likely result in us being right back here shortly thereafter.

Pennsylvania continues to be a state where the pandemic is spreading within the denser metropolitan areas of Philadelphia and Pittsburgh, leaving the central T to see fewer cases that spread more slowly.

Delaware might be approaching an inflection point, given that its most populous county, New Castle, is about to reach 1000 cases. (By the end of today it likely will if its new case trend holds.) We know that deaths lag new cases, and so the worry is that the number of deaths will begin to increase rapidly. The hope is that the slow initial growth of the outbreak will have left hospitals able to better cope over the longer time frame than if everyone had gotten sick all at once.

Virginia is a state that we will contrast to New Jersey, which I will write about last. Because Virginia is a state where it appears the outbreak is beginning to pick up steam and accelerate, rather than flatten. There was the significant drop in cases on Sunday, but that was due to the state’s enhancement of reporting data. (Their website now includes many new statistics.) But just like that the Monday data showed over 450 new cases on Monday. The question will be whether or not over the remainder of the week those new case numbers fall from over 450 to less than 400 to show that the state can flatten the curve before the outbreak becomes especially severe.

Illinois has shown a lot of variability in its day-to-day numbers, hence the advantage of the rolling average. But even that has appeared a wee bit jagged. It’s tough to see the curve flattening just yet, but if we receive updates today and over the next few days that cases are consistently lower, than we might just be able to say the curve is flattening.

And of course in New Jersey we have a state where the curve really and truly has flattened. We have yet to see sustained evidence of a decline in the number of new daily cases. As I said before, this might be more a situation where the outbreak has stabilised and roughly the same number of new cases is being reported daily. Of course the hope is that whatever that rate is falls below the excess capacity threshold of the state’s hospitals.

But I also want to take a look at the state of New Jersey with a degree of granularity. Because, as I noted with Virginia above, not all states are at the same point in their outbreak. And the same can hold true within states. We know that the outbreak in New Jersey began in the north and was very late in reaching some parts of South Jersey. So the same metrics we run for the state, we can run for the counties—though the data I have been collecting from the states only goes back as far as St. Patrick’s Day.

New Jersey's curves — New Jersey’s curves

The northern counties, where the state has been hardest hit, have clearly begun to see the curve bend. But in the south the story is a bit more mixed. Some, like Burlington and Ocean, have seen the curve noticeably flatten. But in Camden and Mercer Counties, home to Camden and Trenton, respectively, the evidence is not quite there. Instead, in these populous counties there exists the very real possibility that the outbreak will continue accelerating for hopefully a very short while.

Credit for the pieces is mine.

Sunday Covid-19 Data

Another day, more cases of coronavirus and Covid-19. So let’s take a look at Sunday’s data as there were some interesting things going on.

First, let’s dispense with Virginia. The state is enhancing its reporting structure, and so they admit the data is likely an underestimate of the present situation in Virginia. So here’s Virginia, nothing really changed.

Moving on, we have Pennsylvania. Here we are beginning to truly see the disparity between the cities in the southeast and southwest, namely Philadelphia and Pittsburgh, and the T that describes what sometimes is used to describe Pennsyltucky. (Though it also includes cities like Harrisburg, the state capital.) The point is that the T of Pennsylvania has yet to suffer greatly from the outbreak. Of course, it’s also the part of the state least equipped to deal with a pandemic.

New Jersey is just bad. One can make the argument that South Jersey is hanging on. (Though I will touch on that later with an idea for today’s afterwork work.) Bergen County in the northeast is likely to surpass 10,000 cases on its own today. And that will put it above most states.

Delaware is tough because it sits as a small state next to several much larger ones. But, the numbers seem to indicate the outbreak is still worsening. Though in terms of geographic spread, there’s little to say other than that New Castle County, home to Wilmington, in the north is the heart of the state’s outbreak.

Illinois is a fascinating state, because of how dissimilar it is compared to Pennsylvania, a state which has a similar number of people.

The map shows that geographic spread still has a little way to go before reaching every county in the state. But the outbreak has been there longer than in Pennsylvania. And most of the darker purples are concentrated in the northeast, in Chicago and its collar counties. Compare that to Pennsylvania above where you will see dark purple scattered across the cities of its eastern third, e.g. Allentown and Scranton, and in the western parts near Pittsburgh. This too could be worth exploring in depth in the future.

Lastly I want to get to the cases curves charts. Here we look at the daily new cases in each state.

The curves, flattening or otherwise, of the five states.

And unfortunately Sunday’s numbers will impact the Virginia curve, but it overall looks as if the state is worsening. I would argue that Illinois, which appears to be bending towards a steadying condition is likely in a weird weekly pattern where it appears to stabilise on weekends and then resumes reported infections come Monday. Pennsylvania might well be flattening its curve. I would want to see a few more days’ worth of data before stating that more definitively. Let’s give it to Wednesday or Thursday.

And then in New Jersey we have a fascinating trend. The curve of increasing number of cases has clearly broken. But it also is not shrinking. Instead, it seems to be more of a plateau. And in that case, the outbreak in New Jersey is not getting worse, but it’s also not getting any better. At least not numerically. However, the goal of flattening the curve is to create a slower, more steady increase in case numbers to help hospitals cope with surge volumes. So good news?

Credit for the pieces is mine.