Support Center

What is the difference between the Wear Time Validation algorithms?

Last Updated: Nov 28, 2016 02:32PM CST

The Wear Time Validation tool in ActiLife allows users to screen epoch-level *.agd files (collected from any ActiGraph device) to flag periods of non-wear, essentially filtering out those periods for further analysis.  Non-wear is estimated by analyzing periods of little or no activity and applying algorithms to those periods to determine if the user was actually wearing the device or not.  Three Wear Time Validation options are available to users:

Troiano (2007)

The Troiano wear time validation option was derived from the SAS code used to filter out non-wear times in the NHANES 2003-2004 dataset.  The original SAS source code can be found at http://riskfactor.cancer.gov/tools/nhanes_pam/  The default settings (Troiano Default) were derived from code located at http://riskfactor.cancer.gov/tools/nhanes_pam/create.pam_perday.sas.  The Troiano algorithm looks for basic patterns in the data and classifies those patterns as wear or non-wear times.

ActiLife offers users the ability to customize the original Troiano settings by making adjustments to the algorithm's spike tolerance, activity threshold, and other settings which can tweak how the the algorithm's detects non-wear time.  For a full description of the parameters, see Troiano Wear Time Validation Parameters.

Choi (2011)

The Choi wear time validation option was derived from Validation of Accelerometer Wear and Nonwear Time Classification Algorithm by L. Choi, Z. Liu, C. Matthews and M. Buchowski (http://www.ncbi.nlm.nih.gov/pmc/articles/PMC3184184/).  In the paper, Choi et.al., compares the Troiano algorithm to a new technique which utilizes a forward and backward looking windowing technique to better categorize non-wear times.  The parameters utilized in the paper can be varied using ActiLife to obtain desired results.  For instance, while the validation relies on only single, vertical axis analysis, ActiLife allows the user to combine all three axes into a single Vector Magnitude (Sqrt[X^2+Y^2+Z^2]) for multi-axis datasets.  For a full description of the Choi parameters, see Choi (2011) Wear Time Validation Parameters.

Floating Window vs Daily Algorithms

Floating Window

Both Troiano and Choi are considered Floating Window algorithms.  These algorithms do not break a file into calendar hours, but rather look at consecutive epochs for patterns.  Users can choose, for instance, to say "only consider 75 minutes of consecutive zeros as non-wear".  After that pattern is detected, all data within that range and the data that follow will be flagged as non-wear, even if that data crosses midnight for a given day.  There are no calendar hour/day constraints to detecting non-wear time.  This algorithm provides the user with much more flexibility by allowing the user to set tolerance levels, minimum wear-time lengths, and non-wear stop levels.  The detailed views and export options do, however, offer the user the option of viewing calendar day specific results as shown below.

WTVDetail.jpg

Daily

The "Daily" option breaks a file down into calendar hours and calendar days.  Essentially, an entire calendar HOUR will be flagged as valid or invalid based on the validation criteria.  At the end of each hour, the analysis starts over.  This option allows a user to view wear time discretely (hour by hour) rather than as indiscriminate time periods.  The Daily algorithm is not based on any existing wear time validation algorithms and was designed by engineers at ActiGraph.  As such, this feature is deprecated - the feature can only be enabled by selecting Tools->Options->Calculations and selecting the "Enable Daily WTV Algorithm" option.  This will add the Daily option to the algorithm dropdown list and will present the user with parameters for identifying hourly wear/non-wear periods.

If you need any further assistance, please visit our support portal at https://help.theactigraph.com/home. From there, you can view many helpful FAQ's and Troubleshooting Articles; log in to check any existing Support Requests; or create a new Support Ticket if you are unable to find a solution. You can also email us at support@theactigraph.com or give us a call at (877) 497-6996, Monday through Friday, 8 am to 5 pm CST.

 

Contact Us

84597daf644fbf443ca13c8a01998d15@actigraph.desk-mail.com
https://cdn.desk.com/
false
desk
Loading
seconds ago
a minute ago
minutes ago
an hour ago
hours ago
a day ago
days ago
about
false
Invalid characters found
/customer/en/portal/articles/autocomplete