Observations, Questions, Concerns:

• The methodology for calculating these projections isn’t disclosed on the graph.
• There are several drops in the actual death counts. These drops may coincide with weekends, when most people aren’t working and data is not reported as quickly to government agencies.
• The most recent IHME projection predicts deaths due to COVID-19 will extend into August--longer than any other projection. We now know it is extending/extended much longer
• There is no epidemiological reason why the data would fit a cubic function, especially since according to these types of models, the number of deaths will increase indefinitely and at some point surpass the population of the United States.

​

​

Context: The Cubic Fit model was originally presented by the White House’s CEA (Council of Economic Advisors) in a tweet from May 5th. It predicts that by mid-May, deaths from COVID-19 would be near 0. Nate Silver from FiveThirtyEight posited that this model came from a pre-canned Excel function that fits data to a cubic polynomial. One key feature of the graphs of these types of functions are that they eventually increase indefinitely.

Observations, Questions, Concerns:

• Deaths have consistently occurring at higher rates among the older populations.
• How many of the older population deaths have occured because of close or communal living in nursing and elderly care facilities?
• If the number of Coronavirus cases are increasing and are currently higher than April levels, how do we explain a consistent reduction in deaths?
• Deaths are roughly at 0 for all age groups for the week ending July 4th as of July 4th, which we know is incorrect.
• Most deaths occurring on July 4th would likely not be reported to the CDC by July 4th. Assuming a delay of at least a few days and possibly a few weeks in reporting, this graph makes it look like deaths are decreasing.
• Numbers are likely higher if people died from the Coronavirus but were never tested for it. Deaths may have been falsely recorded as pneumonia, for example, especially for earlier periods when healthcare providers were not aware that the virus was already spreading throughout the US.

​

​

Context: This graph provided by the CDC gives the provisional number of deaths due to COVID-19 on a weekly basis, based on age group. Data is recorded from death certificates. The data is based only on available data at the time and the numbers for a given week may change in the future as more information is available. The CDC claims that provisional death data is lagged by 1-2 weeks on average.

Context:

This bar chart was created by the CDC and shows the number of Coronavirus cases per age group. The data was retrieved in late June 2020. This chart is no longer available on the CDC website.

Observations, Questions, Concerns:

• The greatest number of cases is among people aged 18-44, followed by people aged 45-64, and then by people aged 75 and older.
• How is this data collected? Is this hospitalizations? Positive tests? Is the reporting voluntary?
• The number of years covered by each bar is different. For example, there are 26 years in the 18-44 group while there are only 9 years in the 65-74 group. This makes it seem like the Coronavirus disproportionately affects young people.

​

​

Context:

The following charts shows the number of confirmed Covid-19 cases in the U.S.over time.

​

Observations, Questions, Concerns:

• The increasing trend is shows in the graph. However, there is no information on “Number of Tests”
• Labels on the x-axis and y-axis are not included.
• The title of the graph is not included.
• The x-axis is labeled inconsistently. At first, every day is labeled. Then, every other day is labeled.

https://www.esri.com/arcgis-blog/products/product/mapping/mapping-coronavirus-responsibly/, retrieved 7/9/2020

Context:

The goal is to represent number of positive Covid-19 cases by location.These graphs are provided in an ArcGIS blog by Kenneth Field.

​

​

Observations, Questions, Concerns:

• The population sizes in each area are different
• The number of cases will depend on the population size.
• The second graph uses a consistent baseline (per 100,000) whereas the first graph concentrates on the number of cases (e.g. >65,000).
• In the first graph, it appears that many regions are affected as severely as Hubei province while in the second graph, it appears that Hubei is affected far worse than any other province.

Context:

This is a screenshot of Milwaukee County Covid-19 dashboard.These numbers are provided daily on the dashboard.

​

​

​

Observations, Questions, Concerns:

• The bar graph presents the racial breakdown of positive Covid-19 cases in Milwaukee County.
• The racial breakdown of positive Covid-19 cases do not include “Hispanic or Latino in Any Race” category on July 2. On July 1, there were 4182 covid-19 positive cases in “Hispanic Latino in Any Race.” On July 3, there were 4255 covid-19 positive cases in “Hispanic Latino in Any Race.”
• Data entry errors might cause problems for researchers.

https://mcoem.maps.arcgis.com/apps/opsdashboard/index.html#/018eedbe075046779b8062b5fe1055bf

Context:

The histogram was originally presented in a blog from Brookings. It was included in an article explaining how the Coronavirus is disproportionately affecting minorities. Brookings used the CDC race/ethnicity definitions, so the white and Black categories exclude Hispanic/Latinos, and Hispanic/Latino refers to Hispanic/Latino of any race.

​

Observations, Questions, Concerns:

• The y-axis should clarify “death rates per 100,000” Also, the y-axis scale does not reflect the “rates” <200 very well
• The data is not adjusted by the demographics of the regions hit hard by the virus, the attribution of deaths to the virus and the population sizes by age groups.
• In every age range, Black people are dying at the same or higher rates than white people who are 10 years older.
• In every age range, both Black people and Hispanic/Latino people are dying at higher rates than white people.

Context:

The histogram was originally presented in a blog from Brookings. It was included in an article explaining how the Coronavirus is disproportionately affecting minorities. It displays the ratio of deaths by race/ethnic group to white deaths. Brookings used the CDC race/ethnicity definitions, so the white and Black categories exclude Hispanic/Latinos, and Hispanic/Latino refers to Hispanic/Latino of any race.

​

Observations, Questions, Concerns:

• The definition of “ratio of death rate” is not included in the graph (though it is described in the article).
• How many graphs are taken out of context and as a result show vague or undefined results?
• 10 times as many Black people aged 35-44 die due to COVID-19 than white people in the same age group. 8 times as many Hispanic/Latino people die than white people in this age group.
• The ratio of Black to white deaths and Hispanic/Latino to white deaths decreases as we look at larger and larger age groups.
• Why are so many more younger/middle-aged Black people dying from COVID-19 than white people of the same age?
• The data is not adjusted by the demographics of the regions hit hardest by the virus, the attribution of deaths to the virus and the population sizes by age groups.

Context:

This graph is provided by Brookings. Both the bar graph and the line graph use age groups on the x-axis, but the bar graph uses the left axis (share of population) while the line graph uses the right axis (crude death rate). Brookings used the CDC race/ethnicity definitions, so the white and Black categories exclude Hispanic/Latinos, and Hispanic/Latino refers to Hispanic/Latino of any race.

Observations, Questions,Concerns:

• Whites make up a much larger proportion of older age groups
• The crude death rate appears to grow exponentially as age increases.
• If COVID-19 affects all races equally, we would expect the deaths due to COVID-19 to occur among the races/ethnicities within each age group in the same proportions as the races/ethnicities are distributed within a given age group. For example, we would expect twice as many white people as black people to die from COVID-19 in the 75-84 age range, since white people aged 75-84 make up 6% of the population but black people aged 75-84 make up 3% of the population. This graph doesn’t tell us this.

Context:

This graph is provided by Brookings in a blog post about how the Coronavirus is disproportionately affecting minorities. Brookings used the CDC race/ethnicity definitions, so the white and Black categories exclude Hispanic/Latinos, and Hispanic/Latino refers to Hispanic/Latino of any race.

​

Observations, Questions, Concerns:

• The COVID-19 death rate is higher for blacks regardless of whether the death rates are age adjusted.
• When the death rates are age-adjusted, it seems the black death rate is over three times the white death rate (3.6x to be exact).
• While there seems to be no difference in crude death rates between Hispanic/Latinos and whites, the Hispanic/Latino age-adjusted death rate is over double the white age-adjusted death rate (2.5x to be exact).
• How are the age-adjusted rates calculated?

​

Observations, Questions, Concerns

• There is an increase in the seven day average of new daily cases in all five states relative to June 1.
• Only in North Carolina does this increase in new daily cases also accompany a comparable increase in daily tests.
• There is an increase of over 150% in the seven day average of new daily cases in Arizona and Oklahoma between June 1 and June 15.
• The seven day average in daily number of tests has decreased from June 1 to June 15 in both Arizona and Oklahoma.
• What would these statistics look like for a blue state over the same period?

Context:

This graph accompanied an article in the Washington Post about how Coronavirus cases were starting to increase more quickly in red states than in blue states . The article appeared on June 17th, 2020.

Context:

This graph accompanied an article in the Washington Post about how Coronavirus cases were starting to increase more quickly in red states than in blue states. The article appeared on June 17th, 2020. These types of graphs are called stacked line graphs.

Observations, Questions, Concerns:

• Over 60% of cases are consistently in states that heavily favored Hillary Clinton (this is not surprising as the largest outbreak occurred in New York City).
• The percent of all cases occurring in states that heavily favored Hillary Clinton has been dropping consistently since the beginning of April.
• The percent of all cases occurring in states that heavily favored Trump has been increasing, but only slightly.
• How many states fall in each of these categories? How many people fall into each of the states in each of these categories?

Context:

This graph accompanied an article in the Washington Post about how Coronavirus cases were starting to increase more quickly in red states than in blue states . The article appeared on June 17th, 2020. These types of graphs are called stacked line graphs.

Observations, Questions, Concerns:

• Over 60% of cases are consistently in states that heavily favored Hillary Clinton (this is not surprising as the largest outbreak occurred in New York City).
• The percent of all cases occuring in counties that heavily favored Hillary Clinton has been roughly constant since the beginning of April. The counties that slightly favored Clinton or Trump have accounted for slightly fewer cases over time. Only counties that heavily favored Trump are accounting for more and more cases (though not by much).
• How many people fall into each of the counties in each of these categories?
• Does this graph with breakdown by county tell a significantly different story from the previous graph with breakdown by state?

Observations, Questions, Concerns:

• Graph shows that initially COVID-19 affected areas that voted against Trump and has slowly shifted to areas that voted for Trump.
• The article argues that the increase in Red states is attributed to different behaviors associated with such political affiliation, namely Red states opened up sooner and do not encourage/mandate mask-wearing.
• This could also be a consequence of where different people of particular political affiliations live in the US and how a virus typically spreads. Urban areas attract liberal voters and are hotspots of viruses because of their high population density. The spread to rural areas, where Republicans live in larger numbers, takes time. Remember, correlation does not mean causation.

Context:

This graph accompanied an article in the Washington Post about how Coronavirus cases were starting to increase more quickly in red states than in blue states. The article appeared on June 17th, 2020. This is a time series and both data points should add to 100%.

Observations, Questions, Concerns:

• This graph records average total cases over three days, but ignores the fact that the population of the United States and the EU differ (the US has about 130 million fewer people).
• The graph states that the actual numbers are higher than the graph shows due to limited testing. Is testing limited in the US as much as it is in the EU? How can we measure this?
• This graph makes it seem like there are far more cases in the United States than in the EU.
• There is a clear downward trend in the 3 day rolling average number of cases in the EU but there seems to be little if any downward trend in the US.
• For both series, there are several drops in the data. Do these drops coincide with weekends, when most people aren’t working and data is not reported as quickly to government agencies?

Context:

This graph appeared on the website for the San Juan Islander on June 20, 2020. In the brief article, it was stated that the chart was released by the European CDC. It also stated that the population for the EU is about 445 million and the population for the US is about 328 million.

Observations, Questions, Concerns:

• The height of the different bars have nothing to do with their values. If anything, the values are almost inversely proportional.
• Was this error intentional?
• There is no scale on the y-axis. We don’t know whether the height is supposed to start with zero or not.

Context:

This graph appeared on a Florida TV station in late June and was passed around twitter to show examples of manipulated graphs. It seems that they updated the numbers on chart, using a visual from previous days.

Context:

These graphs appear daily in the New York Times where they compare the spread of COVD-19 in different countries as well as states. This particular image of just the 4 countries appeared on MSNBC in late June.

Observations, Questions, Concerns:

• The graphs juxtaposed have the same scale along the x-axis, but different scales for each of the y-axes.
• This makes it appear that all the countries had the same number of cases in later March and early April.
• This decision makes it easier to see and compare the change in cases and the rate of increase/decrease. All four countries appear to have the same rate of increase in late March, but the European countries all had decreasing numbers of cases at different rates, whereas the number of cases in the US has stabilized and started increasing.
• Data includes both confirmed and probable cases. How do countries define probable cases? Is this done in the same way in every country?

Context:

The graph shows the “relationship” between physician’s salary and mortality rate of Covid-19 in different states. The graphs secondary source was Twitter, but no original source given. This graph highlights how we need to be cautious of what we see on social media. Posting was dated June 29.

Observations, Questions, Concerns:

• The average physician salary is quite consistent in most states, although the mortality rate for COVID-19 varies considerably.
• Data does not take into consideration cost of living.
• The line of best fit appears to show that as physician salaries descrease so does the mortality rate, somehow implying that paying physicians less would lead to less COVID-19 deaths.
• No correlation coefficient is given. My guess it would be close to zero. There is alway a line of best fit, but it is not meaningful if there is no correlation.

​

Observations, Questions, Concerns:

• The graph shows the number of deaths and hospitalizations resulting from COVID-19 between the dates April 26-May 9.
• Rather than presenting the data chronologically along the x-axis, the author chose to present the data so that the y-values are decreasing. For example, April 26 is after May 7.
• Inevitably, the manipulation of this graph does not build confidence in our government agencies, especially given that this chart appeared 3 weeks after Georgia opened back up and many were critical of Governor Kemp’s decision, arguing Georgia reopened too early.

Context:

This graph appears to show the decrease in cases in all of the 5 most populous counties in Georgia. It appeared on the Georgia’s Department of Health website around May 10. The communications director for Brian Kemp’s office later tweeted that "The x axis was set up that way to show descending values to more easily demonstrate peak values and counties on those dates.”

Observations, Questions, Concerns:

• There is a weak correlation between level of contact and the degree to which the activity is indoors.
• I wonder how they measure these variables.
• Based on the graph, which businesses are safest to reopen?
• Activities that are more dangerous in terms of the spread of the virus would be in the top right corner that are predominantly indoors and require close contact.
• Which activities would you allow to reopen? These more dangerous activities tend to be the more widespread, meaning they have a larger economic impact.

Context:

This graph was presented in a regular column in the New York Times called “What’s Going on in This Graph?” This graph in particular shows the relationship between how much contact occurs between people and how much activity occurs outdoors at specific public locations. The data used for this graph was obtained by having individuals rate how many interactions they had with others, and how much of the activities occurred outdoors at each location.

Context:

This graph accompanied an article in the New York Times on April 2 providing a visual for where Americans were and were not staying home.The article also included a color heat map identifying the amount of travel relative to normal as well as a map color coded to indicate the first date when average travel fell below 2 miles a day.

Observations, Questions, Concerns:

• People living in counties with stay at home orders traveled much less.
• It seems a scatter plot might have shown this relationship more clearly. Why did the author choose to display the data in this way?
• In general, people from densely populated counties traveled much less than those from counties with fewer habitants. Why might this be the case? Might there be other geographical trends, for example travel among people from Northern states versus Southern states?
• Many more counties with large populations imposed stay-at-home orders. This makes sense as rural areas were not affected as severly at this point in the pandemic.
• It is unclear what the coloring represents.

Context:

This chart shows the relationship between preference for Trump and social distancing. Along the x-axis is the percent of population that voted for Trump in 2016 and the amount of decrease in travel over the month of March. This graph appeared in the Economist on April 6th.

Observations, Questions, Concerns:

• There appears to be a relatively strong correlation between how much cities voted for Trump and the amount people are traveling with the latter going up with a higher favorability for Trump.
• No correlation coefficient was given
• While the implication is that voting preference decides how much a person limits their travel, there could be a third explanatory variable. People living in rural areas need to drive more. Moreover, rural areas are more sparsely populated and thus the virus may spread more slowly there. States with more rural areas tend to vote Republican.
• People in Wyoming have actually started driving more since the pandemic.

Context:

This graph, updated daily on Financial Times, illustrates the total number of daily COVID-19 deaths worldwide. In both the top and bottom, the colors show the show the distribution of deaths in different continents. In addition, the top part, the width represents the number of deaths.

Observations, Questions, Concerns:

• Initially (March & April) deaths were predominantly in Europe (Blue) and US (Grey). This has now shifted to South America (pink) and India (red). Africa (light blue) has seen very few deaths.
• The large width shows the maximum number of deaths per day occurred around April 15th. The number of deaths is again increasing since around June 1. Deaths were growing the fastest though at the end of March and early April as the slope is the steepest then.
• I wonder if China is excluded since they are not reporting their cases.
• I wonder why they chose a shape that grows up and down to illustrate total deaths. Wouldn’t the slope be twice as steep if they had used a flat base?
• I notice that North America, Europe, Asia are continents, but Latin America is a larger geographical organization.

Context:

This graph appeared in an Economist article on June 24. On the left, it compares the mortality rate of 5 countries and on the right it breaks up those deaths relative to what percentage fall in different age ranges.

Observations, Questions, Concerns:

• Compared to the UK, Spain, Italy, and France, the US has the lowest overall mortality rate.
• All the countries initially had higher rates of increase in deaths per capita, but the number of deaths in these countries seems to be leveling off.
• The US and the US seem to have the highest increase of deaths per day relative to the others which have slowed down to almost no additional deaths.
• The US has the largest percent of deaths in the younger age groups whereas Europe has had many more deaths in the 80+ range, almost 60% of their totals.
• I wonder why the US contributes to so many younger people dying of COVID-19.The article notes that the US has a lower median age and a higher obesity rate among the middle aged population.

Context:

This graph was produced by Danny Dorling, a UK professor who studies Geography. It appeared in an April 7 article of The Conversation. The graph emphasizes the change in deaths by including it as a variable on the x-axis.

Observations, Questions, Concerns:

• The US has the largest number of deaths (data is not per capita).
• On April 4th, the US death rate had plateaued. The other countries shown seem to have displayed the same zero growth at similar times, although just a few days earlier.
• This growth rate appears on the graph in two places. It is a variable along the x-axis and the slope of the line. That being said, it does make it easier to compare as an actual number.
• I notice that the death rate was consistently decreasing in China.
• What does this graph looks like today?
• How did the author calculate the 7 day average?

Observations, Questions, Concerns:

• The size of the buckets have increased by over 50%
• I wonder if this graph was intentionally created to deceive readers? Who decided to increase the size of the buckets and why?

Context:

These are two graphics of cases per 100K for counties in Georgia as reported by the GA Department of Public Health. The left graph is from July 2 and the second is from July 17. While the graphics look unchanged, the number of cases statewide went up 49% over the same period.

Context:

This graph is presented on APM Research Lab website. The group has been collecting data to present wide disparities by race. This graph shows cumulative actual Covid-19 mortality rates per 100,000 by race and ethnicity

​

​

​

​

​

Observations, Questions, Concerns:

​

• Y-axis and X-axis are not labeled
• Dates are not properly scaled

​

​