Python and the Purkinje

I had an extremely validating experience with a senior researcher the other day. I was at a conference, in the middle of running a series of focus groups, and chatted up a colleague during a lunch break about how fruitful the previous focus group had been. It was my first time running one, and to be honest, I was both thrilled and intimidated by the complexity of data I had collected from a single session.

It’s an interesting tension – knowing that you’ve located a giant treasure trove of messy human insight and patterns, but it’s buried under layers of noise, fragmented and interrupted human speech, and logical dead-ends where I didn’t get to ask “and why is that?”. In a lot of ways, analysis is an archeological process of excavating skeletons by gently brushing away and picking out the dirt in your data, bit by bit. The fruits of this labor can be extremely gratifying, but that warm glow of accomplishment is generally something I feel after the fact. The data collection and cleaning phases that come right before analysis send me spiraling across existential dread, escapist fantasies, and thoughts of a career change. At this stage, I’m essentially facing a mountain of dirt, armed with the faith that I will find something somewhere in that heap. 

I was talking through these emotions with my colleague, and he put words to something I have often felt – the value of resting and taking a break between data collection and analysis. The indescribable benefit of just letting your brain stew in its own juices and work its chemical magic. It’s hard to justify this sometimes – to both myself and my stakeholders. But time and time again, I have reaped the benefits of simply stepping away from my data for bit, and then coming back to it in a few days, if not a week. Over the last three years, after several iterations of my own analysis process and this conversation with my colleague, I landed on a pattern that gives me three large pieces of analysis to do: what struck me, what stayed with me, and what snuck past me. The first piece happens during or immediately after data collection, while the second and third happen after stepping away. 

Data analysis deserves (and will get) several posts of its own, but here’s a brief breakdown of my process, according to these three large analysis phases. The first phase – what struck me – captures my immediate reactions and first impressions of my research and participants. There is a world of information embedded in fleeting gestures, intonations, and pauses that are crucial to capture before they slip away, and I know that I’ll never find them in a transcript or recording. So I find it valuable to capture my visceral, mostly surface-level takes of the conversations I’ve just had. Usually in the form of a debrief and list of bullets, this quick summary serves as a personal polaroid that captures the moment of data collection for me. 

Then comes a break – a few days of stepping away from the data and putting my mind elsewhere. This process seems to naturally wash away a lot of the noise and distill the most resounding patterns in the data – a neurological Principal Component Analysis of sorts. This is the stuff that stays with me, the stuff that participants said over and over again and lodged into my mind, while the rest flowed out with fading memory. These patterns serve as a place to start a deeper analysis of the data. I start with the loudest, most obvious trends and dig around them to find their motivations, constraints, consequences, and blindspots. 

This digging leads me to the stuff that snuck past me. The patterns that sit under the surface, unnoticed by participants themselves. Triangulation across the immediate impressions that struck me at first, the loudest litanies that stayed with me, and a several passes of connect the dots between participants opens up a kaleidoscope of insights that’s only possible when you make different data points face each other and form a picture together. This is the meat of my analysis, and often takes me the longest. But in many ways, it’s also the most fun and intellectually gratifying. 

As I’ve grown as a researcher, this process has gotten faster for me. But I’ve also realized that there is no good substitute for time in the analysis process. While it’s easy to handoff a first pass of data to stakeholders, I personally don’t feel satisfied until I’ve gotten a chance to sit with my own data and wrung out insights. More importantly, the projects where I get to spend a lot of time with my data yield some of the strongest papers because they provide a coherent narrative for my stakeholders. Ultimately, it all comes down to giving myself the time and space to listen to my data and let it tell me a story. And sometimes that means taking a break and listening to something else for a bit.