Charting ResScan and Zeo – for geeks and techies?

How consistent is your length of sleep? It might be better as a percentage if your sleep time varies, that would reduce the noise some.

patrissimo wrote: Can you do a scatter plot rather than having the AI events be cumulative or sorted as it looks like you've done? That would give us a better feel for the correlation - I'm curious what the R^2 on the scatter plot fit line will be

Well, I did do a little work with scatter charts using the variables AI and AVG Deep sleep and AI and Deep Sleep. The result was minimal correlation, r2 of around +0.02 . I also compared the variables Deep Sleep and AVG Deep (the 15 day moving average of Deep Sleep) and the result was only a moderate correlation of r2 =0.45 - I'm not sure I remember all I know about scatter charts, statistics, and the like. In fact, I'm not sure if I remember if I ever knew anything to begin with.
Your suggestion about percentage of sleep time and noise is a good one, though I am a pretty consistent sleeper. Problem is I haven't attempted to clean up this data set. I've got some from both sources that probably should not be included and I have some periods with data from one source and not the other. OK for long term trends and big picture analysis, but obviously noisy.
J

Ok, so I saw in your other Zeo thread that you have very little deep sleep, maybe it's so variable partly because it's so scanty? And maybe there isn't much AHI/Deep sleep connection for you because your deep sleep is getting messed up by something else?

I only have 10 data points with full night CPAP use & CPAP data and Zeo data, but I'm seeing a pretty noticeable correlation, like the R^2 for AHI vs REM% is 0.12 (meaning 12% of the variation in REM % is explained by my AHI), not bad. A new data point every night...I wish I could get 'em faster but I can't .

What vs. What R R^2
AHI vs. Deep % -0.199 0.04
AHI vs. REM % -0.352 0.12
AHI vs D+R-W% -0.045 0.00

Jay Aitchsee wrote:Prior to July 2011, my Deep Sleep as measured by 3 PSG's and Zeo had always been less than 10 minutes

BTW, I was interested to see you say the ZEO matched your PSGs for deep. My last PSG in 2013 showed < 1% stage 3+4, while my 2007 was 9% and 2008 was 10%. Yet my Zeo, for the last few months, has me average 16%, with a 6% standard deviation. So it seems like the Zeo data rates my deep sleep substantially higher, unless it's really improved over the 8 months since my last PSG. It could be my reduction in sleeping pills, I guess.

From the comments and articles I've read, Zeo seems to do pretty well scoring DEEP, not so well scoring REM and Wake. When I compare my daily Zeo reports to daily ResScan and morning/day feel, they seem to match fairly well. I did go back and clean up the data, throwing out short nights, and those that were incomplete. The results are a little smoother. Past charts, over shorter time frames, have shown Deep to be somewhat cyclical without the large day to day variations implied by the first chart in this thread. I'll get another up soon.
In your case, I would tend to trust the Zeo report, at least as a reference for trending. One PSG might be a little suspect. And yes, it could be the pills. Many sleep medications do suppress Deep.
It will be interesting to see how things look after you've gathered more data.
J

OK, so I cleaned up the data (threw out the short nights and orphans) and looked at it again. The result is a little smoother maybe, but that's about it. I did some scatter chart work and, interestingly, found very little correlation between Deep and AI or AHI. I think, since apneas can be spread throughout the night and Deep primarily occurs in the first 120 minutes, that might be expected. I would like to do a correlation between Leaks and Deep, but the same problem exists. Leaks later in the night probably won't have much impact on Deep. Further hampering the Leak/Deep comparison is that ResScan only exports median leak data and both the mean and the mode of my median leaks are 0.00. The 95% leak info might prove interesting, but I don't want to enter it by hand.