Excessive MPA with adult-specific cut points

[Borja Del Pozo Cruz]

Dear Vincent,

I hope you are doing well in these strange circumstances.

I am contacting you because I am an user of GGIR and I think something is wrong. 

When I run the code with cut points proposed for adults by Hildebrand et al. (sed beh 45.8; mpa 93.2; and mvpa 418.3) I have an excessive amount of MPA (like 342 min for some, 142 min for another person....). I feel this is unrealistic. 

Thank you and thanks for making acc data analysis accessible for everyone. 

Best,

Borja

--------

#-----------------------------------------------------# # NOTES # The following notes should be retained regarding the data: # R Version: 3.6.1 # GGIR Version: 1.10-7 #-----------------------------------------------------# #-----------------------------------------------------# # INSTALL # Calls required packages and installs if needed. #-----------------------------------------------------# checkpack <- function(pkg){ new.pkg <- pkg[!(pkg %in% installed.packages()[, "Package"])] if (length(new.pkg)) install.packages(new.pkg, dependencies = TRUE) sapply(pkg, require, character.only = TRUE) } packages <- c("GGIR", "MASS", "signal", "zoo", "mmap", "bitops", "matlab", "GENEAread", "tuneR", "stringr") checkpack(packages) #-----------------------------------------------------# # SETUP #-----------------------------------------------------# # source("file:///C:/Users/s00215420/Desktop/101") #-----------------------------------------------------# # SCRIPT # Runs the script from # https://cran.r-project.org/web/packages/GGIR/vignettes/GGIR.html # with modifications #-----------------------------------------------------# #-----------------------------------------------------# # ***** Child Script ***** # # too include all data processed by GGIR the thresholds have all been set to 0 # these can and probably should be adjusted based off some criteria # i.e., includedaycrit coul dbe 10 hours, includenightcrit could be 4 hours # you may also wish to ignore the first and last day of measurment # you may also want to make adjustments to strategy and def.noc.sleep library(GGIR) g.shell.GGIR( mode=c(1,2,3,4,5), datadir="/Users/antoniogarcia-hermoso/Desktop/accel", outputdir="/Users/antoniogarcia-hermoso/Desktop/accel", do.report=c(2,4,5), overwrite=TRUE, desiredtz="Europe/Madrid", studyname = "Spain", #===================== # Part 2 #===================== strategy = 3, #hrs.del.start = 0, hrs.del.end = 0, maxdur = 0, includedaycrit = 0, qwindow=c(0,24), winhr = c(5,10), qlevels = c(c(1380/1440),c(1410/1440)), ilevels = c(0,45.8,93.2,418.3,8000), closedbout=FALSE, do.imp=TRUE, mvpathreshold =c(93.2,418.3), bout.metric = 4, excludefirstlast = FALSE, includenightcrit = 0, #epochvalues2csv = TRUE, #===================== # Part 3 + 4 #===================== def.noc.sleep = 1, outliers.only = FALSE, criterror = 4, do.visual = TRUE, anglethreshold = 5, timethreshold = 5, ignorenonwear = TRUE, nnights = 9, #===================== # Part 5 #===================== threshold.lig = c(45.8), threshold.mod = c(93.2),threshold.vig = c(418.3), boutcriter = 0.8, boutcriter.in = 0.9, boutcriter.lig = 0.8, boutcriter.mvpa = 0.8, boutdur.in = c(1,10,30), boutdur.lig = c(1,10), boutdur.mvpa = c(1), #save_ms5rawlevels = TRUE, #===================== # Visual report #===================== timewindow = c("WW"), visualreport=TRUE, dofirstpage = TRUE, viewingwindow=1)

[Jairo Hidalgo Migueles]

Dear Borja,

Are you using 5-s epoch and data from wrist-worn devices? I think this is not a specific issue of the cutoff, but of the wrist movement pattern. Do the 1-min MVPA bouts look more realistic? Using this metric you would clean up the sporadic arm movements which are not related to increased PA intensity.

Best,
Jairo

[Vincent van Hees]

I agree with Jairo, another approach to this would be to use 5 second epoch in combination with the bout concept. For example, 1 minute bouts.

Note that the discussion in the literature about the necessity to calculate bouts is confused by a lack of clarify about what we call a bout (= periods of time where we allow for sporadic gaps in behaviour). In my opinion even a 1 minute epoch should be considered a bout, because it looks at a 60 second window and allows for gaps in behaviour within that window. Arguing that bouts are not necessary, would imply that we can process even 1 or 5 second epoch data without using the concept of bouts. Especially in the case of wrist data this is problematic because of the sporadic movements of the wrist/arm.

Another reason why I would argue for using either larger epochs or bouts is that PA guidelines are often monitored with self-report methods, especially in national surveys. Self-report methods are not able to capture short epochs of behaviour, by which accelerometer-derived estimates of behaviour must try to approximate the kinds of (long lasting) behaviours a person is able to recall with self-report. Otherwise accelerometer-based PA research will impossibly be able to guide PA guidelines if we also want to monitor adherence to those guidelines with self-report methods. So, either use 1 minute epochs or use shorter epochs in combination with bout detection. A third option would be to use 5 second epochs with a 1 minute rolling mean applied, but this has not been implemented yet in GGIR.

Vincent

‐‐‐‐‐‐‐ Original Message ‐‐‐‐‐‐‐

--

You received this message because you are subscribed to the Google Groups "R package GGIR" group.

To unsubscribe from this group and stop receiving emails from it, send an email to RpackageGGIR...@googlegroups.com.

To view this discussion on the web, visit https://groups.google.com/d/msgid/RpackageGGIR/09a1cc24-9b7b-4562-9160-b29fe4f681ebn%40googlegroups.com.