Making Maps of a MOOC


This summer I had the opportunity to work with Dr. Anthony Robinson to map MOOC students. Anthony wanted to know more about the geographic patterns of MOOC students in his Maps and the Geospatial Revolution class. We wanted to know such things as: Where are the students? Where are the highly engaged students? Where are the non-English speaking students? Where are students who are employed/unemployed? Are there geographic patterns to the gender of MOOC students?

The goal was to answer all these questions with maps. We wanted to see if we could show the all of the geographic patterns instead of simply saying something like, “X% of students come from the United States”. As cartographers, we believe geographic information is often at its best when visualized on a map.

What’s a MOOC?

It’s a Massive Open Online Course. Anyone can take these courses for free. Check out Coursera (one of the many MOOC providers) to see what you can learn from college professors without putting on pants or paying a dime.

Anthony’s MOOC is a basic introduction to maps and cartography. He keeps it simple for a novices and it has been wildly successful. Over 100,000 people have taken his class in one of the three offerings. That means over 100,000 people around the world are interested in learning about cartography. See guys, maps are cool!

Everyone who signed up for the Maps MOOC

Everyone who signed up for the Maps MOOC

The interesting thing about the students in this type of learning environment is they can technically be from anywhere. All you need is an Internet connection which is easier to come by than ever. We have even seen students who used satellite connections to learn about maps. One cool thing about Anthony’s MOOC is it did not require students to watch the videos, although you are sorely missing out on Anthony’s dry sense of humor without them – the course is designed so students in places with less then perfect internet can still pass the class.

Data, data, data

Coursera, the course provider of Anthony’s class, keeps track of the students who take the MOOC. Researchers can then use this data for multitudes of different purposes. Coursera records the quizzes, page views, forum posts, and video views of each student. Students are also encouraged to take a voluntary survey at the beginning of the class which takes account of their age, gender, occupation, etc. In addition, Anthony had the IP addresses of his students geocoded so we could have detailed locations for everyone who took the class. Combined, we had a wealth of data on the tens of thousands of students all over the globe who had signed up for this particular MOOC.


Mapping this type of data is tough. First, you have to decide on a projection. Since a large portion of the students live in the northern hemisphere you don’t want a projection that distorts that area too much. Anthony was a big proponent of the Goode projection which we both felt reminded us of a 1970s pull-down classroom map. It is equal area, which let us make hexbin maps, and amazingly doesn’t distort most of the world too badly. Goode is Good despite its reminder of burnt orange, brown, and bubble lettering. We hope you like our updated sleek 21st century black, grey, and neon maps with the retro projection.

Everyone who passed the class. Notice our cool colors on the retro projection. #soHipster

Everyone who passed the class. Notice our cool colors on the retro projection. #soHipster

OMG it’s Hexbins guys!

We also really wanted hexbin maps because they are all the rage in cartography right now. Luckily Sterling had already made the hexbins for this project and was kind enough to share with me. As stated above, you should really have an equal area projection when you make a hexbin map because otherwise it totally defeats the purpose of having bins of the same size. Your GIS can help you find an equal area projection, but unfortunately there only a few good ones to choose from especially for mapping MOOC data, where we need to show the whole world at a time.

Hexbin map of everyone who enrolled in the class

Hexbin map of everyone who enrolled in the class

Point Maps

I found in most cases I really liked looking at the basic point maps. Ask a question, e.g. “Where are the people who completed at least one assignment?” Write a query in your GIS and you are done. No need for complicated interpretation or making a legend. These maps are quick and to the point. Literally.

Where are all the students who attempted at least one assignment?

Where are all the students who attempted at least one assignment?

Engagement Index Map

Anthony was insistent upon making a map of an engagement index. Basically this involved figuring out what engagement in a MOOC meant. Since I actually was enrolled in Anthony’s MOOC all three times it was offered, you would think I would know what that means, except I have never gotten past the second week and have failed the course three times. Let’s just say I wasn’t very engaged.

So what did we count into our engagement index?

  •  Video views
  • Forum posts
  • Page views

Using quantiles, I broke each of the categories up into bins. For instance, if you had less than three video views, you were categorized as 1 in the video view category. Between 3-5 video views gave you a 2. Six to 13 video views gave you a 3, and more than 14 video views gave you a 4. I did this for all of the categories (video views, forum posts, and page views), added them up, and that gave us the engagement index, which ranged from 0 to 16. The pattern of engagement is quite interesting.

Where students were more or less engaged.

Where students were more or less engaged.

Bivariate Hexbin Map, now we’re talkin’!

We thought it would be cool to know where people were engaged and where people passed the course. This is most easily done with a bivariate map. Josh Stevens wrote a very informative blog about this a couple months ago. You’ll notice some very interesting geographic patterns in this bivariate map. Three places to focus on here: the US, Europe, and India.

Map of where students had high achievement and were highly engaged

Map of where students had high achievement and were highly engaged

Non-English Forum Posts

We also had data on who was making forum posts, who was making non-English forum posts, and how many posts these people were making. I made a simple proportional symbol map of just that.

Where MOOC students were engaging in forum discussions in languages other than English

Where MOOC students were engaging in forum discussions in languages other than English


It has been known for some time that the majority of MOOC students are male, but we really wanted to know whether this varied spatially. Turns out it does. In the US, more than other places, there is less of a disparity between the genders, however that even varies within the United States. The gender ratio map versus a map of men and a map of women is great for showing where things are more or less equal. You can compare it to the individual point maps of men and women in the class. Which map do you like best? What interesting patterns do you see with gender in these maps?

Gender Ratio Map of MOOC Students

Gender Ratio Map of MOOC Students

Women who were enrolled in the class

Women who were enrolled in the class

Men who were enrolled in the class

Men who were enrolled in the class


Finally, the first offering of the maps MOOC took place in 2013, during the height of some serious economic struggles for many people, especially in places like Spain and Greece. We wanted to see the patterns of MOOC students’ gender and employment status. These two maps are pretty interesting. We mapped employed vs. unemployed men and women. I think Europe is the most interesting especially in Eastern Europe and Spain. There are many more unemployed women taking the class than men especially compared to the employed cohort of each gender. The US has some similar patterns. What do you think might be the cause?

Unemployed/Employed Female MOOC Students

Unemployed/Employed Female MOOC Students

Unemployed/Employed Male MOOC Students

Unemployed/Employed Male MOOC Students

Anthony has been a close friend and mentor for many years, and I owe him a debt of gratitude for how much he has helped me navigate my career; however, we have never actually worked directly together unless you count taking Intro to Cartography with him in 2007. It was the most nerve wracking professional experience of my life. This person thinks you are awesome, and the fear of letting them down is immense. It ended up being fun to get back to my roots with the person who inspired me to become a cartographer in the first place.

Posted in Uncategorized | Leave a comment

Mapping Wind and Population: Using raster algebra and value-by-alpha mapping

I came back to grad school with the plan of studying something that might have a positive impact on the world. When I lived in Southern California, I loved driving out to the desert through the San Gorgonio Wind Farm. It was during one of those many drives when I realized I could make a career out of studying renewable energy. I found it interesting and in studying the geographic/mapping issues around it, I might be able to make a positive impact on the environment. I’m still figuring out how exactly to do that, but in the meantime I have had some time to make cool maps of wind energy.

Wind Farm Picture

Me in the San Gorgonio Wind Farm near Palm Springs, CA

The original goal of this little study was to attempt to find places or a relationship between places where there was high wind and high population. Why? Long story short: wind energy isn’t very useful if it is far from the grid and the people using it. This is where people live:

Population Density

Population Density in the Lower 48 By County

Here is where wind is high:

Wind Speed Map

Wind Speeds in the Lower 48 (Wind data interpolated from 669 METAR weather stations in the lower 48)

Using Raster Algebra to Find Where Wind and Population are High

I converted the population data from polygons to a raster based on my 5-class quantile classification scheme. I then used raster algebra to find where wind and population were both high.

Wind Speeds and Population

Wind Speeds and Population results from Raster Algebra Analysis

The raster algebra map loses a bit of information because it becomes impossible to understand what the middle colors on the legend mean. We can’t tell the difference between places that might have high wind but low population and vice versa.

Using Value-By-Alpha to get a better picture

The raster algebra approach does get me to my end goal of understanding where high winds and high population are collocated, but the technique used on the map below called Value-By-Alpha has been used in other bivariate maps (namely election maps). Basically, the deeper blue, the higher the wind, and the darker, the greater the population. The same general pattern appears, places on the eastern edge of the Great Plains are high in population and also have high winds, but now we can see where high winds and high populations are as well as understand where there might be high wind but low population and vice versa.

Value-By-Alpha Map of Wind Speeds and Population

Value-By-Alpha Map of Wind Speeds and Population

Obviously there are still many limitations to this study, mostly the issue of scale and the lack of understanding of where people actually want a backyard of wind turbines, but those are all issues for another day. In the meantime, I think these maps are beautiful and they offer a better understanding of the geographic relationship between people and renewable energy. Peace, Fish.

Posted in mapping | Tagged , , , , , , | 1 Comment

Booooozy Mapping


I have had plans to start a blog for over two years now and I am finally getting around to it. I’m sure the start of the blog could be its own post, but I will avoid talking about that right now in lieu of having a much more interesting topic.

I have also been planning to go and do my own data collection around the awesome (haha!) Borough of State College, PA (because we don’t have cities in this awesome state). Since this town prides itself on being a “Drinking Town with a Football Problem”, I figured why the hell not make a map of the actual drinking problem?

Now this is actually quite a controversial topic for me. Almost exactly 10 years ago I moved to State College to attend Penn State as an undergrad. While at PSU, I was a good student and a great party girl. I graduated, got my masters, went to work for a while (all stories for another day), grew up a bit, and then decided to make the awesome decision to come back to good ol’ State College for my Ph.D.. Returning to State College was a shock, for lack of a better description. I see the idiotic undergrads walking around like zombies on Friday and Saturday nights and put my head in my hands and think “oh crap, that used to be me”. While I rarely still consider myself a party girl, I obviously am totally embarrassed by the place that first allowed me to become one. But while this has all been going through my head, I have come to realize State College totally does have a drinking problem, and so I figured, “what the hell, I’m gonna put this crap on a map”. So I did.

This past weekend was Halloween, and I desperately wanted to go out and map all the leftover costume bits. Mainly I was trying to locate the elusive Post-Halloween Walk of Shame (perhaps the most embarrassing of all walks of shame). I did that, along with mapping every leftover Solo cup, beer can, and costume bit.


Beer Preferences

So what? All I did was make a map of crap (because that truly is what it is). What if I queried the data with that GIS thing? Alright sure, why not? Here are crappy beer preferences based on my littered beer can survey. Most common beer? Oh don’t worry, we drink high class stuff here in Central PA; it is the one and only Natural Light! “Natty” for the commoners. Bud Light was runner up. Oh you thought we drank Yuengling here in PA? Wrong.

NattyLight BudLight NattyIce MillerLite

Natty Light tends to be concentrated closer to Frat Row while higher quality beer (Bud Light and Miller Lite) are more common in the southern part of the town.

Bag o’ Boooooze

bag natty

Where to go when you want a full can of Natty or an entire “bag o’ booze”? I thought I was the only super smart undergrad when I was here, leaving booze around town to fetch later. Turns out there are a few undergrads just as smart as I once was. One co-ed left an entire duffle bag of booze in a bush. There was also a nearly full case of Natty in another.


A Map of Solo Cups

Why make a map of Solo cups? Because I wanted to make a map of something ridiculous. Really you should be asking: why not?


Halloween Remnants

I wished there had been more Halloween remnants, however, I was not completely disappointed: ribbons, socks, smashed balloons, wig remnants, and broken handcuffs were all left out for the elements to destroy.


Bob Marley was murdered here:


Broken handcuffs can only mean one thing:


Sock Map

On Halloween night you may periodically want to take off your socks and leave them in the street. If you do, I will find them, GPS their location, and put it on a map. Total socks found: 5 (including a matching pair).


Empty Liquor Bottles


When you are too cool for drinking (and littering) a can of beer, you should probably bring an entire handle of Vladmir Vodka (Vlady to us Pennsylvanians). It’s high quality stuff that no self-respecting bar will even dare put in their well. Two words: rubbing alcohol. Undergrads enjoy it neat, with a beer chaser (Natty preferably).

Lower quality stuff continues to be concentrated near Frat Row (Vlady and Bankers), while higher quality liquor is found further west and south.


A GPS App Entirely Composed of Comic Sans!


Alright, so how did I get this data? Well despite it being the day after Halloween, I got my butt out of bed on November 1st, downloaded this terrible iPhone app called Free GPS, walked around for 2.5 hours in the general party area in State College, and literally GPSed every one of the points on the maps above: 299 points to be exact. By the way, Free GPS is an app entirely composed of Comic Sans. One word: awesome. Second word: NOT. Anyways, the app is supposed to let you export your GPS points, but the 299 points I GPSed crashed the app. I ended up typing them all into Microsoft Excel. I obviously had nothing better to do.

The Elusive Post-Halloween Walk of Shame

I will end this post with a final map. I did see it: the elusive Post-Halloween Walk of Shame. The creature, despite the cold, was wearing only shorts and a t-shirt holding a costume in hand. It wasn’t quite what I had imagined. (I really wanted to see someone who was still wearing the costume with smeared makeup everywhere). Oh well, there is always next year…


Posted in mapping | Tagged , , , | Leave a comment