Vizual Statistix

Every year, the USDA’s Pesticide Data Program publishes a report on pesticide residues found on a variety of food products.  These detections, measured on >10,000 samples from across the country, tend to be below EPA tolerance levels.  As such, the products are probably safe for human consumption.  However, it’s still interesting to see which foods, particularly fruits and vegetables, have detectable levels of pesticides.  Unless you’re sure your food is organic, you can use this graph to see which produce has a greater diversity of pesticides, and which items are generally pesticide free.  Products like spinach and strawberries often have a lot of pesticides, and may be better purchased organic. Oranges and grapefruits are usually pretty clean, even when conventionally grown.

Every year, the IRS releases information about the adjusted gross income (AGI) on the individual income tax returns of filers who moved. These returns may have varying numbers of dependents, and the migration data include both in-state moves, where the county of the filer’s address changed, and relocations to a different state. From this, we can get a glimpse into where the money is flowing.

The heat map shows the average AGI of movers from the state on the y-axis to the state on the x-axis, broken into deciles.  Orange shades correspond to wealthier filers, and purple to poorer filers, with the median around $51k. The bar graphs show the weighted average AGI for a given state, based only on interstate mover data; there are many more intrastate movers, and these data would have drowned out the signal from the state-to-state moves. So the bar graph on the top represents average AGI of filers moving to the states labeled on the x-axis, and the bar graph on the right corresponds to filers leaving the states on the y-axis.

On the bottom right, I’ve labeled the top and bottom five values you get from subtracting the values in the top bar graph from the values in the right bar graph. Many people retire to states like Florida, Nevada, and Arizona, so it’s not surprising to see that average AGI is higher for folks moving to these states than it is for those who are leaving. Some of the wealthiest states like New York and Connecticut have the lowest values for this metric, perhaps because people accumulate significant wealth there and then move away. There are also some surprises, like the fact that wealthy people are moving to Wyoming. Overall, the average AGI of people moving within the same state is about $10k less than those moving out-of-state, which is logical given that out-of-state moves tend to be more expensive.

Portland, OR, is embarrassingly white; it is typically listed at or near the top of “America’s whitest cities” lists.  About three-quarters of the population is white.  The lack of diversity is disappointing, and it’s probably not going to change anytime soon.  The city is also very segregated, and gentrification is worsening the problem. This effect is easy to see using census data, but I thought I’d try an alternative approach.

In these maps, I’ve used data from Portland Maps to look at the geographic distribution of home owners’ last names.  I selected four of the most common last names in the city (Smith, Johnson, Miller, and Anderson), and five common last names among Asian people (Chen, Nguyen, Pham, Tran, and Wong). I mapped the locations of the homes owned by people with these last names, and ran a kernel density estimation for each name. Light yellow means there are no homes in the area owned by people with that last name, and dark purple areas indicate a high concentration of people with that last name owning homes in the region.

While the results are not particularly surprising (and don’t account for people who rent their houses), I found one fun fact that caught me by surprise.  While only 7% of Portlanders are Asian, the most common last name of all home owners in the city is Nguyen.  This likely isn’t due to an overwhelmingly large Vietnamese population in Portland, but rather, that a high percentage of people from Vietnam have that name. While it doesn’t fix our diversity problem, it’s pretty cool that there are more folks named Nguyen than Smith!

Before you go to a breeder, please remember that there are thousands of pets at adoption groups waiting for a forever home!  Here, I’ve plotted the distribution by age, size, and gender, of dogs and cats on the website Petfinder.com, which aggregates hundreds of thousands of listings from more than 12k adoption groups.

While you may have your heart set on a puppy or kitten, note that most animals available for adoption are somewhat older. Also, I’m not sure why, but there are many more female cats than male cats listed (about 27% more females), whereas with dogs, the number of males exceeds the number of females by about 9%.

The Jeanne Clery Disclosure of Campus Security Policy and Campus Crime Statistics Act (Clery Act) requires postsecondary institutions that receive federal funding to disclose data regarding crimes that occur on campus. While the act is an important piece of legislation, its impact is severely hampered by the fact that schools fail to provide accurate numbers, likely because the incidents are never reported by the students, or because the schools are not honest in their reporting. The former occurs frequently, and there are numerous well-documented cases of the latter, which have resulted in noncompliance fines.

This issue is particularly problematic when it comes to rape, where underreporting is commonplace. Previous investigations have emphasized that Clery Act data are at odds with more realistic numbers:

http://www.motherjones.com/politics/2015/10/campus-crime-statistics-undercount-sexual-assaults

http://www.aauw.org/article/clery-act-data-analysis/

In this post, I’ve mapped the Clery Act data published in 2015, for incidents occurring during the 2014 calendar year. Overall, more than 90% of institutions reported zero alleged rapes, and less than 1% reported more than 10. I summed the number of reports of alleged rape on campuses by state, then calculated rates per 10,000 postsecondary students. The results are striking: rates of alleged rape are significantly higher in New England than anywhere else in the country. At the opposite end of the spectrum are states like Arizona, Texas, California, and Florida, which host some of the largest universities in the country.

So what is happening? Do Ivy League universities like Brown, Dartmouth, and Harvard (all of which are among the top six schools with the most individual reports) really have more incidents of rape and sexual assault? Or is it that students at those schools are more inclined to report the incidents? Or are the schools more honest in their reporting? Of course, I don’t have an answer for any of these questions, but I find it exceedingly hard to believe that there were fewer than ten rapes at schools with tens of thousands of students like Texas A&M, Arizona State University, Florida International University, and University of Florida.

On a final note, I find it somewhat ironic that the two schools with the highest rates of reported alleged rape are:

1. Pontifical John Paul II Institute for Studies on Marriage and Family: five reports among 64 students, for a rate of 781 per 10,000.

2. Brite Divinity School: nine reports among 196 students, for a rate of 459 per 10,000.

Employees in any role can make missteps – even our highly respected federal air marshals. I came across this dataset, resulting from a Freedom of Information Act request, and thought it would be interesting to visualize. The data represent almost ten years of allegations of misconduct, and the associated outcomes of the 5,213 allegations (a rate of ~1.5 per day).

About the data, according to the Transportation Security Administration:

The data provided in response to the FOIA request reflects allegations of misconduct involving Federal Air Marshals (FAMs) over an approximate ten (10) year period of time from November 2002 to early February 2012. All allegations of misconduct must be reported and investigated pursuant to Agency policy. Therefore, the data includes minor misconduct of an administrative nature, as well as more serious misconduct. In reviewing the entries, it is important to note that the recording of allegations into broad categories of misconduct such as “Absent Without Leave” or “Loss of Equipment” does not necessarily reflect the seriousness of the offense…

To be clear, the vast majority of FAMs are dedicated law enforcement professionals who conduct themselves in an exemplary manner. TSA and FAMS have taken numerous proactive measures to create a workplace model built on professionalism, integrity, and accountability with no tolerance for misconduct…

TSA take all allegations of misconduct seriously. In instances when allegations of misconduct arise, TSA policy requires the prompt and thorough investigation and adjudication of the allegations. As Law Enforcement Officers, misconduct by FAMs is adjudicated by the Office of Professional Responsibility (OPR) which issues adverse, disciplinary, and corrective actions against TSA employees, up to and including removal, for egregious violations that undermine security interests, pose a threat to TSA employees, the traveling public, or result in significant monetary loss. As an independent entity, OPR determines the appropriate level of discipline, if warranted, and holds employees accountable for misconduct. Additionally, all TSA employees, including FAMs, are subject to recurrent annual vetting, including criminal checks and periodic security clearance reviews.

TSA and FAMS continually strive to maintain a culture of accountability within its workforce. Notably, in 2015, as a direct result of internal initiatives, FAMS has seen a significant reduction in misconduct cases resulting in disciplinary actions compared to the time period covered in your request.

Like many agencies, proactive efforts cannot prevent all employee misconduct. There are a handful of employees who betray the trust bestowed upon them. This small group of employees should not adversely reflect on the vast majority of FAMs who are dedicated and committed to performing the FAMS mission to protect the traveling public.

Data source:

Some celebrities like to be seen, others, heard.  I would guess that, as celebrities, many appreciate both. In this graph, I’m inferring their preferences from social media usage, based on how frequently they tweet relative to how often they post to Instagram. (Is this a guaranteed reflection of their feelings? No, of course not. But it’s still interesting.) I’ve plotted this against our preference to hear or see them based on the number of followers they have. Through this, we can explore if we are seeing celebrities who like to be seen, and hearing those who prefer to be heard.

I selected 25 celebrities who are among the top 100 most followed people on both Instagram and Twitter, and calculated how many times more followers they have on one vs. the other. Names on the left side of the graph have more Instagram followers, those on the right have more Twitter followers. The scale shows the number of times as many (i.e., a value of 3x on the Instagram side indicates the person has three Instagram followers for every one Twitter follower).  I also calculated the number of tweets per Instagram post for each person. Everyone tweeted with a higher frequency, but the range was significant: Miley tweets only 1.6 times per photo posted to Instagram, while Ariana’s ratio is 18.3. So I would infer from this that Ariana has a lot to say – she’d rather be heard than seen – while Miley prefers to be seen, given that her ratio is far below the weighted average across these celebrities of 7.1 tweets per Instagram post. Though not shown on the graph, the weighted average for followers is very balanced at 1.07 Instagram followers per Twitter follower.

Sadly, but perhaps not surprisingly, the Kardashian/Jenner family is well followed on both forms of social media; all five daughters have more Instagram followers, with the two youngest having the highest ratio of Instagram to Twitter followers. Only Khloe has an above average tweet rate. Katy Perry, the most followed person on Twitter, and the two One Direction members (Niall and Harry) are all in close proximity within the ‘hear them and be heard’ quadrant. Lady Gaga has a low tweet rate, but has a much stronger following on Twitter than Instagram.

Data pulled on 3/9/16.

Data sources:

https://twitter.com/

https://www.instagram.com/

http://twittercounter.com/pages/100

This simple graphic emphasizes the bias in the USA to live in metropolitan areas, rather than spread out in rural regions. The premise is that I’ve divided the country into ten noncontiguous areas of equal population, with 32 million people each. The areas are grouped based on their population density. In the visualization, the squares are proportional to the land area of each region, with a scale provided.

The results are interesting but not surprising.  The 10% of the USA population that lives in the most rural regions would be spread out across 87.3% of the land area.  On average, you would only find 10 people per square mile in this vast area. On the other end of the spectrum would be a 1,574 square mile region (less than one-twentieth of a percent of the USA land area) that has the population density of Queens County, NY (>20k people per square mile).

All population density calculations were made at the census tract level, using land area only (water area excluded). Fun fact: 96 of the 100 census tracts with the highest overall population density are in New York, three are in California, and one is in Illinois. The one in Illinois (in Chicago) has the highest population density in the country (1,560 people in about three-thousandths of a square mile - 455,643 people per square mile)!

Travelers make claims again the Transportation Security Administration (TSA) for a variety of reasons. The vast majority of these claims (~99%) are for loss or damage of property in checked baggage (76%) or at TSA checkpoints (23%). As it turns out, how that claim is resolved varies considerably from airport to airport.

This graph shows data from 2010-2014 on the 20 airports receiving the most claims, broken down into the three possible decisions: deny, settle, and approve in full. Denied claims receive no compensation. The settlement and approval bars are shaded as a heat map to show median payouts.

SeaTac is the only airport on the list that settles or approves at least half of claims. McCarran International Airport, serving Las Vegas, denies the highest percentage of claims (77%). Median payouts range by about a factor of two. The lowest for settlements is $124.99 (PHX) and highest is $289.00 (ATL). The lowest for full approvals is $54.50 (FLL) and highest is $100.00 (LAX).

NFL and NBA players typically make a lot of money. But in order to get that money, they have to abide by the rules of the leagues. If they break certain rules, they have to pay fines, or lose part of their salaries if they are suspended. As it turns out, NFL players are much more likely to lose money due to suspensions than to fines. In the NBA, the split is about even.

The data in these graphs is based on three complete seasons (2013, 2014, and 2015). The percentages show the money lost for a given type of fine or reason for suspension out of all money lost back to the respective league. For example, 30% of money lost by NBA players is due to technical fouls.

Data source:

http://www.spotrac.com/nba/fines-suspensions/

There are a lot of large clocks in the world. Here are ten of them, plotted to show their relative sizes. As it turns out, Big Ben (ok, that’s actually the bell, but everyone calls the clock Big Ben) isn’t even among the 20 largest clocks in the world. I included it here, though, so you can see just how small it is relative to some of the record holders. Another fun fact: the two largest clock faces in the USA are both octagonal. 

The U.S. Department of State’s Office of Allowances determines the foreign per diem rates for government civilians assigned to official business. These amounts are updated monthly, and any given region can have separate rates for different seasons. The rates are reported for two categories: 1) lodging, and 2) meals and incidental expenses.

I’ve plotted the two rate categories against each other for the 1,044 locations in the dataset. Fifty-five locations are plotted twice for the seasonal rate change. In most cases, lodging allowances are significantly higher than meal and incidental amounts. However, for many of the cities in Venezuela, this is not the case. For example, the per diem for Maracaibo, Venezuela, which is the ninth highest overall ($563), is broken down as $276 for lodging and $287 for meals and incidentals. The highest overall per diem is for Cannes, France in the warmer months ($616), followed by Bermuda ($576). The per diem for Antarctica is the lowest, reported at just $1 (only to be used for meals and incidentals).  Don’t spend it all in one place!

The gender earnings gap is an oft-discussed topic, as it should be. The new College Scorecard website from the US Department of Education offers average male and female earnings by school at six and ten years after enrollment (i.e., shortly after graduation, and after four to six years of work experience). I’ve plotted the change in the gap by subtracting the average female salary from the average male salary for some of the top-tier schools in four categories: medical institutions, private universities, private liberal arts colleges, and public institutions. As you can see, the gap always starts out smaller, but grows as men are promoted and given raises more readily. Out of all the schools, the only ‘negative gap’ – where women had a higher average than men – was six years after enrollment at Haverford.

While browsing games on Steam the other day, I noticed that many gamers who have recorded thousands of hours on a given game don’t recommend it (i.e., they provide negative reviews). This could be for a variety of reasons: they don’t like the game, they liked it but something changed and now they don’t like it anymore, or they do like it but are being funny. Many people change their review from positive to negative due to a glitchy update, charging for mods, increasing the game price (then offering it on sale for the original price), or having too many hackers in the community.

I pulled 3,000+ reviews for each game (1,500+ positive and 1,500+ negative) to create these distributions of hours played by review type. There were two exceptions (Rocket League and Civilization V) where there weren’t enough helpful negative reviews – in both those cases I used all the available negative reviews. The violin plots show the median (white dot), interquartile range (thick black bar), 1.5*IQR (thin black bar), and full range (extent of violin). I’ve limited the y-axis to 5,000 hrs, but many reviews came from players who logged far more – some reviews had over 15,000 hrs on a single game, which is impressive given that a year is ~8,766 hrs.

Of course, the violins show that gamers who recommend a game tend to play it for more hours. That said, the negative review distributions still show significant hours logged! Let’s all try to remember that we play video games because we enjoy them – if they stop being fun, it’s probably time to stop playing.

As it turns out, you can gauge the population density of a city just by looking at the distribution of public libraries.  This, of course, assumes that cities offer roughly the same number of libraries per capita, which may not be perfectly consistent.  Regardless, for these five cities – the largest in the US – the distribution of public libraries does a decent job of outlining the city boundaries and acting as a proxy for the urban density.  No statistical analysis here – the maps are just meant to be visually engaging…

Data sources:

Chicago: https://data.cityofchicago.org/Education/Libraries-Locations-Hours-and-Contact-Information/x8fc-8rcq

Los Angeles: https://data.lacity.org/A-Livable-and-Sustainable-City/Library-Branches/a4nt-4gca

New York City: http://www.nypl.org/locations/

Houston: http://houstonlibrary.org/find-it/find-library-location

Football announcers often mention a player – usually a running back – moving the ball north-south (toward the end zones) or east-west (toward the sidelines). But this is done in the field’s frame of reference, not based on any sort of geographic orientation.

As it turns out, of the 31 NFL stadiums in use, only four (Packers, Broncos, Seahawks, and Buccaneers) are perfectly aligned along a north-south axis (e.g., the sideline or an imaginary line connecting the goalposts would point north-south). Two more are within 5° of north-south (Texans and Bears), and an additional three are within 15° (Rams, Giants/Jets, and Eagles).  Two are perfectly oriented along an east-west axis (Falcons and Vikings).

Of course, there is no rule in football regarding the orientation of the field. Players might say they prefer north-south because of where the sun might be at game time, or what the direction of the prevailing winds is. But most stadiums are often built to fit the available space, which can predetermine the orientation of the field, and for domes, none of the outdoor factors is relevant.

To generate this graph, all azimuths were measured as degrees clockwise from north using Google Earth. For covered stadiums, where the field is not visible from an aerial view, I used seating charts as a supplement. Azimuths were grouped using 10° bins (e.g., 85° to 95°) and then plotted in both directions (e.g., 30° was also plotted as 210°, as fields don’t point in just one direction) so that the graph would have 180° rotational symmetry.

As it turns out, getting pardoned or having your sentence commuted was a lot easier back in the day; now, there is more competition and less leniency. FDR, who granted clemency to thousands, was particularly merciful compared to Obama, who has been one of the least forgiving.