Parse overnight data possible?

[BlueSky1958]

Hey Red - I get your point, but let me elaborate what I see as the difference between "good" bad data and "bad" bad data:

"Good" bad data is if in the normal course of a typical sleep session I experience issues that are NOT attributed to external, uncommon variable. If I'm restless, don;t go into REM (or have issues while in it), or anything else sleep related, that's OK and good to account for as part of what I'd call the "typical sleep session or experience." A crappy nights sleep, barring interruption from other things, is perfectly OK and should be included in an analysis.

"Bad" bad data is like a car alarm going off outside my window at 2AM, or incessant dog barking that prevents falling asleep for along period or waking you up at 4AM, or as happened last night, a moving truck pulling into our cul-de-sac at 5AM and the guys talking loudly. Unplanned, uncommon variables that disrupt what would have otherwise been a normal sleep session (however normal is for you - good of bad).  So if my neighbor decides to jackhammer his driveway at midnight, I need to count my data for that night? I think my cholesterol corollary is a good one to describe how the medical community recommends "adjusting" a variable (fasting, even though it's not typical behavior) in order to get a more accurate assessment (even though you might argue that eating 2 Big Macs and a double cheese pizza every day IS your "normal" diet).

I'm not sure I see the difference between this and providing guidance and instructions on how to change/override machine settings. Anyway, if I've failed to make the case so be it. Just thought I'd throw it out as a suggestion and benefit for users based on my experience.

[Crimson Nape]

Excellent discussion!   A back and forth, causes one to think.    
Keep up the good work.  
- Red

[BlueSky1958]

Okey-dokey! Thanks for listening. Let me know if anything comes of it!

[Gideon]

Phillips threw out some outlier data a bit over a decade ago. That outlier data is what led to the recall of essentially every pap they made prior to two years ago. They literally said this isn't bad enough so we (Phillips) can ignore it.

If that is truly uncommon data then just ignore it. If not then you want to know it, understand it, and react to it, even if your conclusion is to ignore it.

I can see The possibility of shading in some pastel color those areas the user feels are abnormal, possibly displaying the stats for all, current process, and the stats without the indicated area displaying both sets of data. That with some form of notes indicating why the area was flagged.

[BlueSky1958]

That sounds like a different situation than the one I'm presenting. Manufacturer has no role in it. The user alone makes a judgement call if the previous nights data (or segment thereof) represents a valid sleep session and can choose to include/exclude it in the summary stats. All th data remains on the card. A toggle allows to include or exclude prior to running a report and whether it is included in the summary stats. They chan change their mind later if they wish. Easy peasy breezy.

I think there's a risk of making this too complex. It's a simple option that would allow a user to say "last night's sleep sucked since my neighbor decided to work on his car in the driveway all night long - nobody could sleep with that" and omit that segment of data from their stats/report (provided its a separate data log segment and not part of a much larger segment in which you cannot isolate those hours). If two days later you change your mind you can go back to that days record and flip the switch to include that data in everything.

In my attached example you may see why - In my first segment I had an AHI of 1.04 which is very consistent with recent readings. But after I was woken up by the sound of my dog puking, My second segment came in at 9.52, entirely inconsistent due to the late time, jumping out of bed to move her from the carpet to the tile, and after being awake for about 15 minutes. I want to omit that segment. Unusual circumstances.

Something like:
1. From the Daily view - select the segment of data (color bar in lower left corner) that you wish to omit from being counted in the stats or a report you run
2. From top menu bar: Data >Advanced>Purge Current Selected Segment>Message: "Omitting this data from Reports and Stats does NOT delete the data. This option is reversible"
3. Add the option to disable to include the data if desired

[Gideon]

I'm saying flag it unusual circumstances and present both sets of results. Then you can easily compare the differences.

I have had individuals say I don't need to see the full chart, actually insisting on it. After a week of arguing he posted the full set of data and after a 15 second look at his full set of data I solved his problem.

The bottom line is I feel it is essential to include all the data. After you show me that view, and by all means tell me that you fell the first two hrs or whatever are junk because of a, b,c. I'll add that input into my consideration.

[BlueSky1958]

Not sure how to interpret your reply so I'll just say there's no issue with presenting both sets of data, or allowing the user to toggle between views, and to allow your "flagged data" to remain flagged so you can see it in the trend over time.