Log in

No account? Create an account

Previous Entry | Next Entry

~Thursday~  So, it's been a little over a month now since I started using my Fitbit, and as those of you who know me well know, I can't go much longer without doing something with the data it's collecting on me.

I'm pretty sure I've worn the device every day, but there were five nights during this period that I didn't wear it to sleep.

I really bought this device to help keep track of my walking, but I did find the sleep-tracking functionality that it offers most intriguing, as there were two things that I wondered about my sleeping habits that I thought this device might help me explore:

  1. That I fall asleep fairly quickly, and before this device I would have said, "90% of the time, I fall asleep within five minutes of hitting the pillow."

  2. And, I would have said, "I average between 5 and 6 hours of sleep per night."

So, do the statistics support these perceptions that I had?

As you can see, my average time to fall asleep is 7.5 minutes, so I overestimated how quickly I fall asleep. With that said, however, I am incredibly grateful for how relatively quickly I do fall asleep, as I have several friends on Facebook who routinely deal with insomnia, and it really sounds exasperating. I particularly love the four data points of 5 minutes or less.

Minutes to fall asleep; average is 7.5 minutes

I was actually pleased to find that I underestimated the number of hours of sleep I get each night on average, but it's still not where it needs to be. I really need to work on reducing the number of data points that fall below the 6-hour mark.

Hours slept; average is 6.2 hours per night

Of particular interest is this next set of data, which was something I'd not really thought of before I got this device. If you had asked me how often I wake up, on average, during the night, I would probably have answered, "Three, or four times at the most."

This was how I would have described it: "Most of the night I sleep on my back. I would estimate that I wake up three times during the night, once to roll over onto my left side, always to that side first, and then I fall back asleep right away. After a couple hours I wake up and turn 180° to my right side, always that side next. And then, a couple of hours later, one final time to turn back on my back for the rest of the night. Oh yeah, and there's a good chance I'll get up to pee during that third wake-up."

Well, the reality of it all is a little startling, as you can see. I wake up, on average, eleven times per night! However, since my "sleep efficiency" (which is shown in the last graph) is consistently high, I must fall right back asleep after most of these calculated times "awakened."

Number of times awakened during the night; average is 11 times

And finally, there's this cloaked-in-secrecy "sleep efficiency" number that gets reported, and whose algorithm for calculation is not explained anywhere by the Fitbit folks, nor anywhere else with authority on the Internet that I've been able to uncover. There are several "theories," however, but I'm not too interested in people's speculations.

To that end, I just use it as a "relative" indicator. In the 90's seems good, and as long as my average stays in that range, I'll pretty much ignore this indicator. With that said, I would welcome a clarification from Fitbit about what it's supposed to be telling us. I mean they don't have to tell us how it's calculated, per se (But why not?), but at least provide some heuristic as simple as, "91-100% is healthy," 81%-90% is healthy, but you should seek to improve," and "80% and below may indicate a potential sleep disorder, such as sleep apnea or insomnia."

Sleep efficiency (for relative comparison only, as Fitbit does not disclose how it calculates this number; average is 93%

So, what's a guy to do? Here's a place to start: I need to rethink a lot about the time I get to bed. It's all over the map—or the clock, as it were: [Click calendar image to enlarge.]

Times at which I went to bed on the nights in January that I tracked

Times at which I went to bed on the nights in January that I tracked


( 5 comments — Leave a comment )
Feb. 17th, 2013 11:13 am (UTC)
sleep efficiency
Not so top secret! It's just the % of time spent in bed when you are actually sleeping. For you it should be high since you fall asleep quickly.

There are some interesting studies of how accurate these various machines are--search on google scholar for the papers. There are probably some measurement issues with all of them.

Schlaf gut!

Feb. 17th, 2013 06:11 pm (UTC)
Re: sleep efficiency
It's close to that formula, but there is another factor in there. Some examples:

Time Asleep Mins (A)Time in Bed in Mins (B)(A)/(B)Efficiency % Fitbit Reported

So, there's something else at play there.

Edited at 2013-02-17 06:20 pm (UTC)
Feb. 17th, 2013 06:27 pm (UTC)
And you can see in this one...
379 minutes asleep divided by 402 minutes in bed, which comes out to 94.2%, But the sleep efficiency number is 95%.

Edited at 2013-02-17 06:29 pm (UTC)
Feb. 23rd, 2013 12:50 pm (UTC)
Looks like some sort of error correction to me. Look how it's always UP. I'll bet you they know they have a measurement error problem and they just factor it up somehow to compensate. The question is whether it's consistent error or not. If it is, you can compare over time and not worry about the scale.

The fact that they don't tell anyone how they calculate it should be a big red flag to anyone who thinks this widget is truly accurate. Lack of transparency on formulas and methods is not a good thing.

Ann (data dog)
Feb. 23rd, 2013 05:47 pm (UTC)
It may be an error correction, but...
...I also wondering if it's not some factor of how many times you woke up.

In any case, you're absolutely right about the lack of transparency on formulas and methods raising red flags!
( 5 comments — Leave a comment )



Latest Month

February 2017


Powered by LiveJournal.com
Designed by Paulina Bozek