Seeing things differently
By Pip Karoly
The other day I learned an important lesson about why it pays to get creative and spend some time playing with data.
My research is in understanding brain waves - especially when things go dangerously wrong, like they do during a seizure. A seizure is when the brain’s rhythms become too big and too synchronized instead of carrying useful information. Anyway I’d been looking into patterns of seizures and another smaller kind of synchronized brain wave called a spike-wave discharge.
Average spike rate and seizure rate plotted over the course of a day for 15 patients
So we published these results (above) looking at what time of day spike-waves and seizures were happening. It’s really interesting that there are circadian patterns in the data, and the patterns seemed to be patient specific.
After taking a course in the online data visualization tool, d3 (run by Isabell @Isa_Kiko and Rob @robrkerr), I got kind of addicted to playing around with different ways to look at data.
Average spike rate and seizure rate plotted over a 24 hour period for 15 patients
When I used d3 to plot all the patients spike rate and seizure rate around a circle (above) I noticed something new. There is a definite ‘hot spot’ in spiking activity in the late afternoon (increased red in the left hand circle around 12 - 5pm), and also a real dip in seizures in the early morning (greener area between midnight at 5am in the right hand circle).
I have some theories about what’s going on (if you’re interested there’s a video here), but the moral of this story is that looking at the exact same information in a different way can lead to brand new ideas.
And it’s not just me - you may recognize the map (below) as one of the most famous examples of data visualization leading to new insight. It’s a plot showing the geographical spread of cholera cases in Soho, London during an 1850s epidemic.
The map of all Cholera cases recorded by John Snow | Photo courtesy Wikimedia Commons
The creator of this graphic, John Snow
No not that one..
The real John Snow was an epidemiologist, and based on this data he showed that cholera cases were highly concentrated near water pumps. He is credited with debunking the theory that cholera was spread through the air, and showing that it is transmitted through contaminated water sources (there’s a longer blog post here).
Have I convinced you yet that pretty graphs are important? In case not, here’s one last data tale…
You might have noticed in my first figure (the bar plot showing spike rate and seizure rate) that I doubled up on the x-axis (i.e. the time covers two days not just one). That was inspired by this very cool plot taken from a sleep study done in the 1970s.
Raster-time showing sleep cycles, reproduced from Czeisler, et al. “Timing of REM sleep is coupled to the circadian rhythm of body temperature in man.” Sleep 2.3 (1979): 329-346.
Sleep is as tricky as seizures when it comes to understanding what on earth is going on in the brain and why. Some groundbreaking experiments in the 1970s involved removing people from time (by living in isolation from mainstream society) and observing their sleep rhythms, as well as temperature, hormone levels etc.
A weird thing happens to human sleep when our body clocks are allowed to ‘free-run’ in this way. It’s called internal desynchronization.
What happens is the amount of time you sleep becomes messed up - people can stay awake for 48 hours without realizing, and go to sleep for some epic amounts of time too. Scientists couldn’t figure out why this was happening. You might expect that the longer you stayed awake, the longer you would sleep, but that wasn’t the case. So what was the cause of these weird, long sleep cycles?
Two scientists (Elliot Weitzman and Charles Czeisler) cracked the code with a clever bit of data-viz. They plotted sleep time as a raster plot over two days in a row (above). They noticed that the long sleeps (long bars in the picture) and short sleeps (short bars) lined up along different parallel diagonal lines (i.e. different phases of the same cycle). They discovered this was linked to circadian temperature regulation. Our temperature fluctuates very consistently throughout the day, and what we know now is that sleep time depends on your core body temperature at the time you fall asleep. (If you want to read more - I first read about that sleep study in the book Sync by Steven Strogatz).
Anyway the point I’m trying to make is DON’T feel bad if, like me, you procrastinate for five hours making a single plot. You are actually searching for deep scientific insight.
