The Good, the Bad and the In-between

(or How Random is Random?)

I’ll admit my bias as a visually oriented person. I am personally drawn to Trixie’s Sleep Maps for their intricate, cartographic patterns. However, while these charts are excellent at expressing disorder — a single glance reveals the unpredictable chaos of the last 7 weeks — they are not so good at revealing patterns, except in cases where the pattern is overwhelming, such as the vertical columns of the ideal sleep map.

The underlying problem is that unless a sleep map is awfully close to ideal, it’s always going to appear chaotic. These maps are good at revealing either perfect or imperfect situations; they have a hard time teasing out any middle ground.

Given this bias, how do you judge how bad a sleep map is? How do you decide what’s a little bit random and what’s a lot random? Even if part of the map stabilizes, such as Trixie’s overnight sleep habits did for the last week, how can you evaluate the rest of the chart?

This problem can be solved if the sleep data is converted into another format. When expressed as a scatterplot, all sorts of very specific patterns are revealed – including Trixie’s infamous ~38 minute cat naps – as well as more subtle trends and even anomalous incidents.

The Distribution Charts below reveal that there are more patterns at work here than previously thought:

Scatterplot: Distribution of Sleep

The most prominent feature of the Sleep Distribution Chart is the almost solid black line near the bottom that runs from about 7 in the morning to 7 at night. This cluster of dots represents the infamous 38 minute (+/- 5 minutes) naps that are so infuriating to deal with. Fortunately there is a new trend developing directly above this area that represents longer, more substantive naps ranging from 1 1/2 to 2 1/2 hours. We look forward to this line darkening in. We also hope to see more plots in the upper right-hand corner which reflects Trixie going to bed between 9pm-midnight and sleeping for 7+ hours.

Analysis of Patterns and Developing Trends

Scatterplot: Distribution of Wakefulness

The Wakefulness Distribution Chart does not have as many defined regions. The chart could almost be described by a single arch stretching from early morning to late night that reaches it’s apex around 4:30pm. This reflects her tendency to stay awake progressively longer periods as the day progresses, and then shorter periods from late night to early morning. The main area of concern on this chart is the medium density region between midnight and 6am. These points indicate where she has woken up in the middle of the night, sometimes for 30 minutes or more.

We now understand a little more what we are dealing with. Yet the question remains – Are we any closer to cracking The Sleep Code?

This entry was posted in Metrics, Sleep. Bookmark the permalink.

3 Responses to The Good, the Bad and the In-between

  1. FrumDad says:

    Okay, this tears it. I’m now moving your contact info from “T,” for Trixie Update, to “S,” for Super-Geek/Dad (such that the “super” modifies both).

    And as a new father, let me add my voice to the cheering chorus when I see “New Trend: sleeps through the night.”

    (I realize the post to which I’m responding is from January, so the “hooray” is a little late. But I just found this site a little while ago, and I’m going through the archives.)


  2. badgerbag says:

    You are a supergenius! Edward Tufte would be very proud.

    Everything you are writing about and all your analysis is exactly what the world of “baby books” desperately need. I’m not even a guy, and most baby books made me barf with their sappy, mealymouthed, overly cutesy or folksy condescension. Bring on the data and the geeky humor! Hip hip for information density!

  3. Tiffany says:

    Really good work.
    I found a lot of profound information which can help me to go on.
    Thanks for all this input.