5,800 years of data from The Metropolitan Museum of Art

One a cold Tuesday in February, The Metropolitan Museum of Art quietly released data and images on it’s entire collection to the public.  With over 200,000 pieces in it’s collection, The Met is the largest Museum in the Western Hemisphere, and contains relics from 3,800 B.C.

I’m a member of the Met, and try to visit ever 5-6 months.  While I enjoy the experience, one concerning theme that I noticed with this dataset was the lack of data governance.  While it’s understandable that certain pieces would be missing information due to age and lack of record keeping, I found lot’s of objects missing basic data.  In some cases, when the person categorizing the data wasn’t sure, they added a “(?)” after the name.

Country of origin

79% of the pieces in the collection don’t have a country listed

57% don’t have an artist name (not including objects attributed to anonymous)

10% don’t have a classification (ie: Print, Drawings, Ceramic)

The Met has more than paintings

  • 6.8% of pieces are silk
  • 4.5% are etchings
  • 3.7% are Photos

Top Artists at the Met

  • 57% of pieces don’t have an artist listed
  • 2,908 pieces from Allen & Ginter (mostly cards from a tobacco company)
  • 314 Rembrandt
  • 24 Van Gogh
  • 23 Pollock
  • 16 Michelangelo

Conclusion

When I heard that the Met had released it’s data to the public I was excited because I though this was an opportunity to find interesting facts and trends on the different pieces.  What I found was missing datapoints, and inconsistent data that made the dataset difficult to navigate.  I think my next option is to throw this into Python and clean up the data.  I’ll continue to look into the data more and hopefully will have a better post in the future.

Data Sources:

The Metropolitan Museum of Art via Google Bigquery 

Microsoft Excel for visualization

Become a Met Member

Analyzing 99 Million Taxi Trips Using Chicago Open Data

The City of Chicago released a dataset containing 100M trips over four years and it’s a huge win for the Open Data community. In this post, we examine the dataset which tells us everything about a passenger’s journey through Chicago, and see dive into the data to see how the industry is beginning to decline as competition from “Ride Shares”, begin to enter the market.

What’s does the average Chicagoan Tip their taxi driver?

Typically, 21% and that’s been stable since 2013.  Riders do seem to be more generous in December, with 2015 and 2016 having an average of 22% or higher.

Average Monthly Tip % from 2013-2016
What do people normally tip?

In statistics, a dataset has a “normal distribution” when the mean = median = mode,  in normal terms that means the average = middle number in a dataset = number appearing the most.  From the graph above, we see that it doesn’t have the smoothness of a bell curve, but instead, has sharp spikes around the values of 0%, 20%, 25% and 30%.

From my experience with New York Yellow Taxi Cabs, I assume that the payment system presents passengers with a predefined tip amount when paying.  Based on my analysis, 39% of all passengers use a predefined amount, with non-tippers making up 7% of all rides, and tippers (using the 20/25/30 amount) making up 32% of rides.

Note: Tip amounts in the dataset were only available for passengers who used a credit card.

 How are passengers paying for their ride?

There has been a steady increase in the amount of taxi rides that use a credit card.  From January 2013 to December 2016, the amount of trips using a credit card has increased from 30% to 47%.

Fewer people are using taxis.

2016 was the worst year for the Chicago Taxi industry with only 19.8M rides, the lowest in four years, and 26% lower than 2013.  Interestingly this didn’t have a huge effect on total fares, while trips were down 26%, fares are dropped 18%.  Similarly, while trips are down 10% from 2015, fares have only dipped 1%.

Putting on my economics hat to investigate this decrease, one possible reason for the decrease could be that riders are substituting taxis for ride sharing apps like Uber or Lyft which provide the same service at an equal or lower price.  Or perhaps it’s the January 2016 fare increase of 15%, that has driven consumers away.  In fairness, a 15% increase in Price and a 10% decrease in Quantity would suggest that the demand is slightly inelastic, but I digress. Other less likely reasons could be changes in public transportation, bikeshare programs, or more walking.

How fast does a taxi travel?

How long does a passenger spend in a taxi?

I found an interesting trend in average trip duration which seems to follow a seasonal trend. The winter months tend to have a shorter trip duration.  With a brutal winter, many passengers likely opt to take a taxi for shorter distances than they would in summer months.

Average time a taxi was in service

Average Monthly Fare

Similar to the average trip duration, the average fare follows a similar trend.  Winter months have a lower overall fare than the summer months.  January seems to be consistently 10% lower than May for each year in the dataset.

 

Conclusion

The Chicago Taxi business is in decline and has seen +10% decreases for two consecutive years.  The introduction of competition from ride share apps like Uber and Lyft has surely eaten into their business and will continue to increase market share as their businesses expand.  From a data perspective we found interesting stats on tip percentage, speed, and ride duration while also witnessing the affects of weather on how the city commutes.  This was an interesting dataset, and I want to look closer into the effects of how neighborhoods pickup/ dropoffs, but first I have to learn about Chicago neighborhoods (or do they have buroughs like NYC?).  My next steps will be to compare NYC to Chicago to see how each city’s taxi compares to the others.

Sources:

Big thanks to the City of Chicago (and more specifically Freedom of Information Laws) which released taxi ridership data into the public domain, and another bigger thanks to Google for adding this into the Big Query dataset and allowing users access to query the information and use it for free.  Chicago, NYC and several other datasets can be found in their database.

Data Sources: Taxi Data via Google Big Query

Code and queries: I need to set up a Github link with code used to generate these queries.

Visualizations Sources: Graphs created using Microsoft Excel.

There are 10,665 people in America named Shaq

During Super Bowl XLI, you may have seen a commercial about “Super Bowl Babies“,  which are babies conceived immediately following a city’s Super Bowl win.  While I believe the link between conception and championship is difficult to prove, it doesn’t mean that sports can impact a parent’s decision making.  This commercial did remind me of a Clemson football player named Shaq Lawson, which makes him the second person named Shaq I had ever heard of.

Using the Social Security Administrations data, and Google Big Query, I decided to look at the names of Hall of Fame players with unique names, and look at the number of babies born during that time with their name.  I chose five athletes, Shaquille O’Neal, Michael Jordan, Tiger Woods, Kobe Bryant, and Lebron James.

 

When did each name peak?

 

Shaq (’92, ’93, ’94)

Before 1973, Shaq (and Shaquille) did show up in the SSA’s published data, but twenty years later peaked at 2,422.

The most popular years for the name Shaq began after Shaquille O’Neal was drafted out of LSU with the first draft pick in the 1992 NBA Draft.  During his first three years with the Orlando Magic he won Rookie of the Year, appeared on the cover of Sports Illustrated, and finished fourth in MVP voting.  ESPN listed Shaq as the fourth best center of NBA history.

Kobe (’00, ’01, ’02, ’03)

The first name Kobe went from 307 in 1996, to 1,093 in 1998 which coincided with Bryant’s second season with the Lakers.  From ’01 – ’03 Bryant finished in the top ten in MVP voting each year, and won three NBA championships during that time.

 

Jordan (’90, ’91, ’97, ’98)

Jordan was next, there were 660 people named Jordan in 1977, but 22,080 in 1990, and was by far the most popular of the five names analyzed.

The top years for the name Jordan coincided with Michael Jordan’s peak with the Chicago Bulls.  During his tenure, Jordan won six NBA championships and Finals MVPs, and 5 All-Star awards.  Jordan is considered to be the best player in NBA history.

 

Tiger (’97, ’98, ’10)

Tiger won’t go down as the most popular name, but it is a unique first name that first appeared in 1997 after Tiger Woods won the 1997 Masters Tournament.  Interestingly enough, the most popular year for the name Tiger was 2010 when Tiger was the center of an infidelity scandal.

Lebron (’07, ’10)

I was surprised to see the Lebron hasn’t been a popular name despite James’s championship wins with the Heat, major endorsement deals with Nike.  Only two years have had more than 50 people named Lebron born: 2007 and 2010.

I recorded every step I took for over two years

I used my iPhone to record every step I took from November 2014 to December 2016.  I found some interesting trends after looking at average daily temperature, day of week, and location.

Some like it hot

After downloading the data from my phone, I compared my steps to the daily average temperature (source: NOAA) to see how temperature affected my activity.  On ‘hot’ days (temperature higher than 83 degrees), I averaged more than 11k steps.  This is because I like to take advantage of the warm weather and run 3 – 6 miles along the East River park on weekends, and squeeze in a couple of 2 – 4 mile runs on weekdays.  What’s interesting is that there seems to be a baseline step amount of 8,700 steps, meaning, that no matter the temperature, I always needed to walk 8.8k steps.  I cover this later, but this was likely because of my apartment’s distance from the subway.

The graph below is is probably my favorite representation of temperature and frequency.  I found this graph type when I was looking at a DC Bike Share on Kaggle.com.  On the x-axis is the temperature, the y-axis is my steps, the colors represent the type of weather (cold vs hot) and each marker repersents a day over the last 2+ years.  Like the previous image, you can see a slight increase in daily steps as the temperature rises.

NYC Real Estate

From 2014-2016 I lived in three different places, Connecticut, Manhattan, and Brooklyn.  My most active days were when I lived in the East Village, where the closest subway was almost a mile from my apartment.  Just getting from my apartment to work, meant that I walked almost two miles roundtrip. Things changed when I moved into a new apartment in Brooklyn.  This apartment was much closer to a subway, and my steps reflect it.  Keeping all things equal (steps taken on the way to get lunch, running/gym) we can assume that I take 2,000 fewer steps a day.

 

I took a look at the average amount of steps I took each month.  During the spring/summer months we notice that I took over 10,000 steps!

What hours did I walk the most in 2016?  To no one’s surprise, I walked the most during my weekday commute 9am, 6-8pm.  As mentioned earlier, I walked almost 1 mile to the nearest subway which would account for the 800+ steps I took during those times.

 

How do you track every step you take?

First you’ll need an iPhone with the Health app installed, which allows your phone to track every step using movement, similar to a pedometer.

Next, you’ll need to download this app: QS Access which will allow you to export a CSV of your daily steps onto Google Drive.

PIVOT TABLE Tutorial for Beginners

Pivot tables are a powerful tool for data analysis in Excel.  Large datasets can be easily digested in a fraction of a second to provide you with more relevant and actionable information.

 

Imagine a data file with every purchase you made last year. How would you calculate how much you spent at at Starbucks?  What about the average amount? Without pivot tables, you would filter the data, then copy it into another tab, and then use a sum on it.  With a pivot, you could select Starbucks and your total expenses in three clicks.

Step by Step Instructions

  1. Highlight the cells you want to pivotCreate a pivot table in Microsoft Excel
  2. Open the “Insert” tab which is located towards the top of the Excel window
  3. Click the Pivot Table button
  4. When the Pivot Table window pops up, click “ok” to create a pivot table in a new tab

Calculated Fields

Pivot tables have a secret feature that allows you to create an entire new field in the pivot table which can help create ratios using two columns from your data.   In the image below, I’m calculating the average fare of a taxi ride in New York 

City by dividing the Fare (Sum of Fare) and dividing it by the Trips (Sum of Trips).

 

examples of a calculated field
Digital Advertising Calculated Fields

Click Through Rate = Clicks/Impressions

eCPM = 1000 * Ad Revenue/Impressions

 

CPC = Ad Revenue/ Clicks

taxi Calculated Fields

Tip % = Tips/Fare

Average Fare = Fare/Trips

 

How to run 300 miles

In January 2016, I set a goal for myself that I would run 300 miles in 2016.  This would be a 50 mile increase from 2015 – a year where I ran more than any year previous –  and an 80 mile increase from 2014.  The rules were simple, only running counted toward 300, of the miles run, they had to be recorded in the app, MapMyRun, warmups and cool-downs don’t count towards the mileage, and treadmill running doesn’t count unless it’s over 1.5 miles.

Setting Quarterly Goals

I knew from the start that it’s unlikely that I’ll run the same every week – temperature, daylight, and likelihood of injury all factor into my running decisions.  With this in mind, I set the following quarterly mile goals:  50, 75, 125, and 50 miles.  My rationale was simple, I have less of an opportunity to run in Q1 & Q4 because of weather and an early sunset.  Injury is a major concern too, if I go out to fast in Q1, I could injure myself and be out for a month.

My July – September goals were justified because of the same reasons as my Q1 goal.  From May to August, the sun sets after 7:30p, which allows me enough time to leave work and run 3+ miles.  This takes the burden off of weekends, which means that I’m not dependent on 2 out of 7 days for my ten mile weekly goal.  Also, longer days mean I don’t need use the gym treadmills where I run 2.5 miles before getting too bored to finish.

Reality vs Goals

Part of this post is to be honest with myself.  Even though I modeled this down to the week, I still fell well short of my goal.  To avoid failing again, we need to learn from my mistakes.

We can see that my actual running in Q1 fell short of what I thought was possible.  This is the result of several factors: weather and work.  Jan and Feb were pretty brutal, there was a “blizzard” the first week of February which prevented my from running most of Feb.

Q3 was the biggest reason I missed my goal.  In January I set a personal goal of 125 miles over a 13 week span.  In theory this was easy to achieve, I would need to run 10 miles a week, or run 3.33 miles three times a week (probably Wednesday, Saturday and another date).  In reality I wasn’t able to meet this goal, for reasons I’m still trying to understand.  Part of it may be related to my personal life, I went on more dates on weekdays and weekends.  Another reason may be the 3 mile races I competed in every Wednesday.  These races were tough, and I felt like I had to rest for several days to recover from the races.

Next Steps

Since I didn’t run 300 miles last year I’ve decided to fulfill this goal in 2017.  I’ll need to learn from the mistakes of 2016, and be more proactive in Q3 when the goal is higher.  I’ll also need to determine if I’ll accomplish this with a lot of short runs (1.5 – 3 miles), or longer runs 4 miles.  I’ve entertained the idea of training for a half-marathon to accomplish this with less runs, but I prefer to keep my training (and racees) under 6 miles per run.