How to Make GGIR Use Sleep‑Log–Defined Wake‑Up Time as start_end_window in Part 5 (WW)

[huilynelleshi-cmyk]

Dear GGIR team,

I am new to using GGIR and I am currently processing ActiGraph GT9X raw data for 24‑hour compositional data analysis. My goal is to use sleep duration derived from the sleep log, combined with daytime behavior measures from the GGIR output.
I am using the sleep log as the guider (with relyonguider = TRUE and sleepwindowType = "TimeInBed"). However, I noticed that in Part 5 (WW), GGIR still uses wakeup_ts as the start_end_window, rather than guider_inbedEnd_ts. My intention is for GGIR to use the sleep‑log–defined time‐in‐bed window (guider_inbedEnd_ts) as the start/end of the 24‑hour window in Part 5. Could you please let me know whether this is possible, or whether GGIR is designed to always use wake‑up time as the anchor for Part 5?

I also have two questions regarding sleep log formatting:

Basic Sleep Log:
Can I include separate In Bed Date and Out Bed Date columns in the basic sleep log format, or does GGIR require only clock times (with dates implicitly inferred)? Our actigraphy protocol collects 10 days of wear data, but after manual inspection of the sleep diaries, we excluded some nights as invalid for certain participants. We would like GGIR to avoid processing those invalid nights, so I want to confirm the best way to specify this in the basic sleep log.

Advanced Sleep Log (Calendar‑Date Format):
Do the D*_date columns in the advanced sleep log represent the calendar date (as described in the documentation)? For example, suppose Participant A begins wearing the monitor at 5:00 pm on 2/10/2021, goes to bed at 10:30 pm on 2/10/2021, wakes up at 7:33 am on 2/11/2021, and then goes to bed again at 1:00 am on 2/12/2021. In this case, would the advanced sleep log be correctly specified as:

D1_date = 2021‑02‑10, with D1_inbed = 10:30:00 and D1_wakeup = NA
D2_date = 2021‑02‑11, with D2_wakeup = 07:33:00 and D2_inbed = 01:00:00
… and so on?

In our study, participants started wearing the device on the afternoon of Day 1 (around 5 pm) and removed it on the afternoon of Day 11. Therefore, for D1_date, we typically have an in‑bed time but no wake‑up time. When I tested the advanced sleep log, I noticed that GGIR used the second wake‑up time from my sleep log as the first guider_inbedEnd_ts, instead of using the first wake‑up time. I would like to confirm whether this is expected behavior due to the calendar‑date structure of the advanced sleep log, or whether this indicates an issue with how I formatted my sleep log input.

Thank you very much for your time and guidance. I want to ensure I understand the intended behavior of GGIR before proceeding further.

Any clarification would be greatly appreciated.

Kind regards,

Hui

[huilynelleshi-cmyk]

Dear GGIR team,

I am following up on my previous questions. I have now resolved the formatting issues for both the Basic Sleep Log and the Advanced Sleep Log, and those inputs are working correctly.

However, I still need help with my main question:

How can I make GGIR fully use the sleep‑log–defined wake‑up time as the start_end_window in Part 5 (WW)?

The sleep log we provide to GGIR (dates, wakeup time, and inbed time) is derived from ActiLife using the Cole–Kripke algorithm applied to count‑based ActiGraph data. We have used these ActiLife‑derived sleep estimates in several published papers, and for consistency with our previous work, we would like Part 5 of GGIR to use the sleep‑log–defined wake‑up time as the start/end anchor for the 24‑hour window, rather than GGIR’s internally estimated wakeup_ts, even though the two values are very close.

Because our 24‑hour compositional data analysis depends on the ActiLife sleep log defining the day boundary, I would like to confirm whether GGIR allows Part 5 (WW) to rely entirely on the sleep log (i.e., guider_inbedEnd_ts) for defining the 24‑hour window, and if so, how to configure this.

Thanks again,

Hui

[vincentvanhees]

I have now updated the code to also allow for activity diary based segmentation in combination with WW, MM and/or OO timewindow:
https://wadpac.github.io/GGIR/articles/TutorialDaySegmentAnalyses.html. There is a new column window_name to clarify which day segment a column corresponds to.

Could reinstall and try again please?

[huilynelleshi-cmyk]

Hi @vincentvanhees,

I would be happy to help test this again. However, I do not see the new window_name column in the activity log structure on my end. Could you please send me a screenshot showing how to add window_name to the activity log?
Also, just to confirm, should I replace qwindow = c(0, 8, 12, 18, 24) with qwindow = "Advancedlog.csv", Is that correct?

Thanks,
Hui

[vincentvanhees]

It should be all the way on the right end of all the columns, did you do search with ctrl + F for it?
Does GGIR generate the segment reports?

Also, just to confirm, should I replace qwindow = c(0, 8, 12, 18, 24) with qwindow = "Advancedlog.csv", Is that correct?
That is correct

[huilynelleshi-cmyk]

Hi @vincentvanhees,

I searched for window_name, but I was not able to find it. I did find qwindow_name. Is qwindow_name the same as window_name?

Once I figure out how to add window_name to the Activity Log, I will run a test analysis.

Could you also update the GGIR tutorial to include more information about window_name, if it is needed for running day-segment analyses? I think that would be helpful for users encountering the same issue.

Thanks,
Hui

[vincentvanhees]

I have removed column qwindow_name, because the segment names are already part of the column 'window'.
There is no need to add window_name to the Activity Log, the Activity Log remains the same to how it has been for years.

Yes, the documentation will be updated to reflect this new functionality.