See also complementary vignettes on: General introduction to GGIR, Cut-points, GGIR parameters, Embedding external functions (pdf), and Reading ad-hoc csv file formats.

1 Introduction

Is this specific person more active in the morning or in the afternoon? Are children more active during their work hours or their leisure time? Questions like these can be answered with GGIR but you first have to specify a few parameters.

The main input argument to be specified is qwindow, which can be used the following ways:

  • To specify the clock hours in the day based on which the segmented day analyses should take place.
  • To specify an activity log (diary) that should be used to guide the segmentation per individual and per day of the recording.

In the following sections I will discuss both scenarios.

2 Clock hour-based segmentation

To perform clock hour segmentation, you will need to provide function GGIR with argument qwindow and assign it a numeric vector with the hours for the segmentation. If the start and end of the day, are not explicitly provided in the vector GGIR will add them. Please find below some example values for qwindow. The number of values used by qwindow is unlimited, but be aware that some of the analyses are such as MX-metrics are impossible for very small windows and will produce empty results.

qwindow value Resulting segment(s) to be analysed
c(0,24) midnight to following midnight (24 hours), the full day is the only segment.
c(8,24) midnight-8:00 (8 hour segment), 8:00-midnight (16 hour segment), and midnight-midnight (24 hour segment).
c(6,11, 13, 17) midnight-6:00 (6 hour segment), 6:00-11:00 (5 hour segment), 11:00-13:00 (2 hour segment), 13:00-17:00 (4 hour segment), 17:00-midnight (7 hour segment), and midnight-midnight (24 hour segment).
c(0:24) 25 segments: 24 segments of 1 hour corresponding to each hour of the day, and midnight-midnight (24 hour segment).

Day Saving Time (DST) is taken into account when identifying the start of the day, but not when identifying the day segments. In other words, a 23 hour days is processed as the 24 hours after the first midnight. This to ensure that segment length is identical across days of the week, which is needed to ease comparison of outcome variables across days.

3 Segmentation guided by activity log

To perform activity-log based segmentation, you will need to provide function GGIR with argument qwindow and assign it the full path to your activity log in .csv format, e.g. qwindow="C:/myactivitylog.csv".

The activity log is expected to be a .csv-file with the following structure:

ID date work travelhome home date work travelhome home
1234 04-11-20 7:45:00 17:00:00 17:30:00 05-11-20 17:30:00
4567 24-11-20 7:45:00 17:00:00 17:30:00 25-11-20 7:45:00 17:00:00 17:30:00

Rows: First row represents the column headers after which each row represents one accelerometer recording.

ID-column: The first column is expected to hold the recording ID, which needs to match with the ID GGIR extracts from the accelerometer file. If unsure how to format the ID values, apply GGIR to a sample of your accelerometer files using the default argument settings. The ID column in the generated part 2 .csv reports will show how the participant ID is extracted by GGIR. If no ID is extracted, see documentation for argument idloc, which helps you to specify the location of the participant in the file name or file header. If ID extraction fails the accelerometer files cannot be matched with the corresponding activity log entries.

Date-column: The ID column is followed by a date column for the first log day. To ensure GGIR recognises this date correctly, specify argument qwindow_dateformat. The default format is "\%d-\%m-\%Y" as in 23-2-2021 to indicate the 23rd of February 2021. If your date is formatted as 2-23-21 then specify"\%m-\%d-\%y". The column name of the date column needs to include the character combination “date” or “Date” or “DATE”. Use the same date format consistently throughout your activity diary.

Start-times: The date column is followed by one or multiple columns with start times for the activity types in that day format in hours:minutes:seconds. Do not provide dates in these cells. The header of the column will be used as label for each activity type. Insert a new date column before continuing with activity types for next day. Leave missing values empty.

Missing values: If values are missing the preceding and succeeding time point will be used as the edges of the segment. In the example below this means that we will define a segment from A-C for ID 1234, while for ID 6789 we only defined segments A-B and B-C, segment A-C is not derived here.

ID date A B C
1234 04-11-20 7:45:00 17:30:00
6789 24-11-20 7:45:00 13:00:00 17:30:00

Notes: - If an activity log was collected for some individuals then those will be processed with qwindow value c(0,24). - Dates with no activity log data can be skipped, no need to have a column with the date followed by a column with the next date. - The end time of one activity is assumed to be the start time of the next activity. We currently do not facilitate overlapping time segments.

4 Examples

For more information about how to use the GGIR function call see explanation in the main GGIR vignette.

4.1 Clock-hour based segmentation:

GGIR(datadir = "/your/data/directory",
             outputdir = "/your/output/directory",
             mode = 1:2, # <= run GGIR parts 1 and 2
    = 2, # <= generate csv-report for GGIR part 2
             qwindow = c(0, 6, 12, 18, 24))

4.2 Activity log based segmentation:

GGIR(datadir = "/your/data/directory",
             outputdir = "/your/output/directory",
             mode = 1:2, # <= run GGIR parts 1 and 2
    = 2, # <= generate csv-report for GGIR part 2
             qwindow = "/path/to/your/activity/log.csv")

After running this code GGIR creates an output folder in the output directory as specified with argument outputdir. In the subfolder results you will then find three files:

  • part2_summary.csv the recording level summary, with 1 row per recording and in recording level aggregates of day segments in columns.
  • part2_daysummary.csv the day level summary, with 1 row per day and day segment specific outcomes in columns.
  • part2_daysummary_longformat.csv the day level summary in long format, such that each row represents one segment from one day in one recording.

In both part2_summary.csv and part2_daysummary.csv the column names tell you the day segment they correspond to. For example, column names ending with _18-24hr refer to the time segment 18:00-24:00. In part2_daysummary_longformat.csv the time segment is clarified via columns qwindow_timestamps and qwindow_name.

5 Analyses performed per day segment

The analyses that GGIR per segment of the day, include:

Acceleration distribution: Derived if argument ilevels is specified. You will find these under the variable names such as [0,36)_ENMO_mg which means time spent between 0 and 36 mg defined by acceleration metric ENMO.

Number of valid hours of data: You will recognise these as N_valid_hours_in_window which tells yoyu the number of valid hours per time window, and N_valid_hours which is the number of valid hours per day.

LXMX analysis: LXMX analysis, which stands for least and most active X hours of the segment. You will recognise these variable names like L5hr_ENMO_mg which is the start time of the least active five hours defined by metric ENMO, and L5_ENMO_mg which is the average acceleration for those hours.

Intensity gradient analysis: You will find these as variables that start with ig_gradient_ See description of GGIR part 2 output in the main GGIR vignette for further details.

Time spent in Moderate or Vigorous Physical Activity (MVPA): You will find these as variables such as MVPA_E5S_T201_ENMO or MVPA_E5S_B1M80%_T201_ENMO. See description of GGIR part 2 output in the main GGIR vignette for further details.