impute_missing_data: Impute missing data

In arctools: Processing and Physical Activity Summaries of Minute Level Activity Data

Impute missing data

Description

Impute missing data in minute-level activity counts data vector based on "average day profile".

Usage

impute_missing_data(
  acc,
  wear_flag,
  valid_day_flag,
  imputeFromValidDaysOnly = TRUE
)

Arguments

acc	A numeric vector. A minute-level activity counts data vector. It is assumed to be in midnight-to-midnight format, meaning its vector length is a multiple of number of minutes in a full day (1440; see midnight_to_midnight()).
wear_flag	An integer vector. Wear/non-wear flag (1/0) for each minute of activity counts data. It is assumed to be in midnight-to-midnight format, meaning its vector length is a multiple of number of minutes in a full day (1440). See midnight_to_midnight(), get_wear_flag().
valid_day_flag	An integer vector. Valid/non-valid day flag (1/0) for each minute of activity counts data. It is assumed to be in midnight-to-midnight format, meaning its vector length is a multiple of number of minutes in a full day (1440). See arctools::midnight_to_midnight(), arctools::get_valid_day_flag().
imputeFromValidDaysOnly	A logical scalar. Whether or not data from valid days only should be used for computing "average day profile" used for imputation.

Details

An "average day profile" is computed as average across minutes identified as wear and from valid days (see param. imputeFromValidDaysOnly). Activity counts data are imputed from "average day profile" for minutes identified as non-wear in days identified as valid, except for minutes before/after data collection start/end which remain NA.

Theoretically, it is possible that all valid days of data collection have non-wear flag for the some minute(s) (i.e., somebody is always taking off the watch for the same few minutes during a day) so there is no data to use to compute imputation values from. If it happens, then method uses 0 as imputation value(s).

Value

A numeric vector. A minute-level activity counts data vector with data imputed for minutes identified as non-wear in days identified as valid

Examples

## Read exemplary data
fpath_i <- system.file("extdata", extdata_fnames[1], package = "arctools")
dat_i   <- as.data.frame(data.table::fread(fpath_i))
acc     <- dat_i$vectormagnitude
acc_ts  <- lubridate::ymd_hms(dat_i$timestamp)
## Get acc data vector in "midnight_to_midnight" format
acc <- midnight_to_midnight(acc, acc_ts)
## Get wear/non-wear flag
wear_flag <- get_wear_flag(acc)
## Get valid/non-valid day flag
valid_day_flag <- get_valid_day_flag(wear_flag)
## Impute missing data in acc data vector
acc_imputed <- impute_missing_data(acc, wear_flag, valid_day_flag)
## Compare mean acc before/after imputation
c(mean(acc, na.rm = TRUE), mean(acc_imputed, na.rm = TRUE))

arctools documentation built on Nov. 11, 2022, 1:05 a.m.