I'm not sure this is the place to discuss it, but I saw several comments about removing morpheuz in favor the brand new Pebble Health sleep tracking. As far as I'm concerned, I'm still using morpheuz for the smart alarm and for the data export.
However, I was surprised by how good is the correlation between the amount of deep sleep reported by Pebble Health and how rested or sleepy I feel throughout the following day.
I have had trouble correlating data from morpheuz with how I feel, so I'm asking here whether there is a way to improve it (or improve my processing of its data).
What I found the most surprising with morpheuz results from my nights is that there is very little "light sleep" and a very high frequency of transitions. The larger part of each of my nights consists of a deep sleep point, followed by a restless/awake point, followed by a deep sleep point, and so on. I have seldom two points in a row in the same category (except in the restless/awake category on bad nights and lazy mornings), and I don't think have 3+ consecutive points in deep sleep more than once or twice a week.
The end result is that usually I have a single-digit percentage of light sleep, and about as much deep sleep as restless/awake. And the ratio of deep sleep to restless/awake look much more like statistical noise than an indication of rested/sleepy.
Before Pebble Health, I thought that was just how my sleep was, I tried hard to make sense of morpheuz data. But now Pebble Health has at least shown that there exists some way to use the data to get meaningful results for me.
So the kind of graphs I get look like severe undersampling. But then again, you're using "max" as a summary function, and it blows any anomaly out of proportion. However, I can't imagine what kind of anomaly there could be, especially since I tested with another unit with similar results, which makes a hardware defect unlikely. My first instinct would have been to use RMS instead of max, so that larger measured values still have more weight but a few anomalous measurements are still drowned in correct data.
Anyway, you seemed to have spent a lot of time researching sleep tracking, so I thought it would be more productive to start a discussion instead of experimenting on my own.
I don't mind moving to a different platform if such a discussion should not happen here, and while I don't want to publish my sleep data to the world, I can share it privately if it helps.