For a long time I’ve been fascinated by the Quantified Self movement. As someone passionate about data, I can’t think of any prettier data than the kind related to myself. That’s why, since the fall 2013 I began to collect data about myself using different kinds of software and gadgets that track my life seamlessly.
Now, the time to start digging into this data has arrived, especially since I realized yesterday about Tableau’s Quantified Self Viz Contest. Though I’m planning to build more complex, advanced and “cross-topic” visualizations, I decided to give a second try to Tableau after my series’ visualization. The dataset that I’m exploring in this post is related to the number of keystrokes that I perform on my laptop, my main working tool. The data has been collected with this open source keylogger which provides a file with a timestamp and number of keyboard strokes performed by minute.
To start with, my job is to analyze and visualize data, hence most of my work is about coding and using softwares such as R and Excel. Therefore it has to be understood that keyboard strokes are not the only metric to have a look at in order to understand my working habits. Nevertheless the dataset is large enough to enable spotting common things such as lunch time located around. Apart from that, as expected, a general decrease in the amount of time spent using my keyboard can be spotted as the weekend gets closer. Moreover, and surprisingly to me, it looks like in the case of having to work during the weekend, I do it mainly on Sundays (perhaps I tend to do more reading on Saturdays if I have to).
In a more artistic way, the bottom visualization looks like a music spectral analysis and shapes the way I use the keyboard by locating every single keystroke record per minute in the weekday pane and at the height related to the number of strokes
With the development of this visualization I’ve learned two things about Tableau:
It is a very easy tool to use in order to explore possible visualizations and prototype visualizations
It is complicated to work with visualizations that render thousands of items (such as the spectral visualization), as any change done implies a large computational time. Maybe it could be useful to have a “don’t render” option that could freeze the details of the visualization while one can change other properties such as axes and titles.
Visualizing data is not about showing pretty pictures, but a way to learn from the insight they provide. In that sense, this has been my first attempt to make sense of all that keystrokes I’ve been recording for the last few months. It is indeed important to stress that this data does not fully reflect my working habits, but it can certainly point me to some behaviors that could be deeply studied when combined with other data sets.
P.S.: I have been unable to align the line charts in the middle of the dashboard with the “spectral visualization” at the bottom. This is because I have no records about keystrokes between 0 and 6am on Mondays and the “Show Missing Values” is not adding them. Any help on that will be much appreciated!
UPDATE: Thanks to Johanna Schlereth for the suggestion on the alignment problem! It was a matter of setting the x axis of the “Aggregated keystrokes per weekday and hour” to continuous, and the set the axis to “fixed”. Thank you!