Non-wear time and timewindow

[Malontax]

I am new in using the GGIR software for accelerometer data and some principles remain a bit unclear for me.

• If I have do.imp = TRUE enabled and 27% of a participant's time is classified as non-wear, but the day still meets the requirement of having >16 hours of valid data, is the entire non-wear period imputed with the average of the other days of that week?
• Do the absolute minutes, such as dur_day_total_IN_min and dur_day_total_MOD_min, include only minutes spent at that intensity, or can minutes of non-wear be classified in these variables?
• If my research focuses on physical activity during the day, which window is best for my analysis: "WW" or "MM"?

I hope my questions are clear and that you can help me with them.

[vincentvanhees]

If I have do.imp = TRUE enabled and 27% of a participant's time is classified as non-wear, but the day still meets the requirement of having >16 hours of valid data, is the entire non-wear period imputed with the average of the other days of that week?

Yes, they are imputed based on time of the day. The default of 16 hours is fairly low, but chosen in the context of studies where participants are asked to wear the accelerometer 24/7. Ideally the 16 hour criteria is only relevant to catch the days with really poor compliance, while for most days the participant still wears it for much more than 16 hours. So, if you see that a person consistently does not wear the accelerometer for a substantial part of the day but still more than 16 then that may still be a reason to exclude it or to treat it with extra caution.

Do the absolute minutes, such as dur_day_total_IN_min and dur_day_total_MOD_min, include only minutes spent at that intensity, or can minutes of non-wear be classified in these variables?

Non-wear is imputed and after the imputation treated the same as data points classified as wear. This is needed, because we ultimately want to make statements about typical behaviour, which is not possible if we would leave gaps in time. As a result, dur_day_total_IN_min, dur_day_total_LIG_min, dur_day_total_MOD_min and dur_day_total_VIG_min add up to the total duration of the waking hours in a day.

If my research focuses on physical activity during the day, which window is best for my analysis: "WW" or "MM"?

Personally, I think "WW" or "OO" are best because only they ensure that the waking hours of a day are treated as one window regardless of whether that window overlaps with midnight. If a person falls asleep after midnight window "MM" will combine that wake time between midnight and sleep onset with the activity during waking hours the following day, which to me is non-intuitive.

That said, there are researchers who prefer to work with 24 hour windows via MM. This is why GGIR facilitates all of them.

[Malontax]

Thank you for your reply and explanation. It clarifies things much better.

I have another question: Is it correct that when you set do.imp = FALSE, possible non-wear is classified as inactivity instead of being replaced by the average activity at that time, as it would be if it were set to TRUE? Does setting do.imp=FALSE therefore possibly cause your total inactivity to be higher?

For my research, I would like to show the participants' compliance or average wear time, but I am having difficulty determining which variable to use for this. This is because the non-wear percentage included in part 5 data is not the only factor, as all days with very low compliance have already been excluded.

[vincentvanhees]

I think I have now answered this in #1470 (comment)