R/process_accel.R
exclude_accel.RdThis function subsets accelerometry data by total wear/non-wear time criteria, as well as NHANES data quality flags.
exclude_accel( act, flags, threshold_lower = 600, rm_PAXSTAT = TRUE, rm_PAXCAL = TRUE )
| act | Activity data matrix. Should contain, at a minimum, the columns: SEQN, PAXCAL, PAXSTAT, WEEKDAY, SDDSRVYR. In addition, the activity data matrix should have 1440 columns with MIN1, MIN2, ..., MIN1440. |
|---|---|
| flags | Wear/non-wear flag matrix. Should contain the same columns as the activity data matrix, in the same order. However, instead of activity counts, the columns MIN1, MIN2, ..., MIN1440 should be binary (0/1) wear/non-wear flags where 1 indicates wear and 0 indicates non-wear. |
| threshold_lower | Lower limit on the amount of wear-time for a day to be considered good. Defaults to 600 minutes which implies a "good" day has at least 10 hours of weartime. |
| rm_PAXSTAT | Logical value indicating whether to remove based on the data reliability indicator variable PAXSTAT. Defaults to TRUE. |
| rm_PAXCAL | Logical value indicating whether to remove based on the accelerometer calibration indicator variable PAXCAL. Defaults to TRUE. |
This function returns a numeric vector containing the indices of days which were identified as "good". These indices can be used to subset the accelerometry data as desired. An illustration is provided in the examples.
Fundamentally, all this function does is check to see if a day has a user specified amount of estimated total wear time (ignoring timing of wear) and checking for whether any data quality concerns were indicated by NHANES via the PAXSTAT and PAXCAL variables. Because this function doesn't actually use the activity count data, it's not technically necessary to include the activity data matrix. However, forcing the activity data matrix to be included and checking that subject-days are identical between the activity and wear/non-wear flag matrices adds a layer of protection against subsetting or sorting one data matrix but not the other.
This function ignores missing data. Missing values to not count toward (or against) wear time.
Hart, Teresa L et al. “How Many Days of Monitoring Predict Physical Activity and Sedentary Behaviour in Older Adults?” The International Journal of Behavioral Nutrition and Physical Activity 8 (2011): 62. PMC. Web. 10 Oct. 2018.
library("rnhanesdata") ## remove all days with fewer than 10 hours of wear time in the 2003-2004 accelerometry data ## and exclude all days with data quality/calibration flags data("PAXINTEN_C") data("Flags_C") ## obtain indices for "good" days using the default threshold of at least 10 hours of weartime keep_inx <- exclude_accel(act = PAXINTEN_C, flags = Flags_C) ## subset the accelerometry and flags data using these indices accel_good_C <- PAXINTEN_C[keep_inx,] flags_good_C <- Flags_C[keep_inx,] ## check that all remaining days have at least 10 hours of wear ## and there are no data quality issues as flagged by the PAXSTAT and PAXCAL variables summary(rowSums(flags_good_C[,paste0("MIN",1:1440)], na.rm=TRUE))#> Min. 1st Qu. Median Mean 3rd Qu. Max. #> 600 754 854 875 950 1440#> #> 1 #> 32806#> #> 1 #> 32806