Using the IntakeBalance Package

## Generate the result
vector_result <- IntakeBalance::IntakeBalance(
  fm_start = 50.5,   ## Fat mass at the start of the measurement period
  fm_end = 50.7,     ## Fat mass at the end of the measurement period
  ffm_start = 75.1,  ## Fat-free mass at the start of the measurement period
  ffm_end = 74.9,    ## Fat-free mass at the end of the measurement period
  fm_units = "kg",   ## Units of measure for the fat mass value
  ffm_units = "kg",  ## Units of measure for the fat-free mass value
  ee_per_day = 1950, ## Daily energy expenditure
  ee_units = "kcal", ## Units of measure for the daily energy expenditure value
  n_days = 14        ## Number of days between the start and end measurements
)

## Show the result
knitr::kable(vector_result)

## Create the data frame
info <- data.frame(
  PID = c("001", "002"),
  fm_start = c(50.5, 70.2), fm_end = c(50.7, 70.0),
  ffm_start = c(75.1, 90.3), ffm_end = c(74.9, 90.3),
  ee = c(1950, 2473), n_days = c(14, 14)
)

## Show the data frame
knitr::kable(info)

## Generate the result
df_result <- IntakeBalance::IntakeBalance(
  
  ## These arguments still refer to the same information outlined previously,
  ## but now we have added a layer of abstraction to reference the names of
  ## columns where that information is stored, rather than providing the
  ## values themselves
  fm_start = "fm_start",
  fm_end = "fm_end",
  ffm_start = "ffm_start",
  ffm_end = "ffm_end",
  ee_per_day = "ee",
  n_days = "n_days",
  
  ## The trick is to pass in a data frame via this extra argument. That's how R
  ## knows to interpret the other variables as column names rather than raw values
  df = info
  
)

## Show the result
knitr::kable(df_result)

## Find the file
filename <- system.file(
  "extdata/TAS1H30182785_2019-09-17.gt3x",
  package = "read.gt3x"
)

## Read the file using the `accelEE` package
ex_data <- accelEE::ee_file(filename = filename, scheme = "Hibbing 2023")

## Extend the data file for the sake of illustration
## (ignore this part when processing real-world data)
ex_data <- data.frame(
  Timestamp = ex_data$Timestamp[1] + (60*0:3199),
  do.call(rbind, replicate(80, ex_data[ ,-1], simplify = FALSE))
)

## Show some of the contents
knitr::kable(head(dplyr::select(ex_data, !matches("features"))))

## Set a starting point for the random number generator so that sampling is
## repeatable when generating the `Axis1` and `valid_status` columns
set.seed(610)

## Now generate the sleep/non-wear dataframe
agd_data <- data.frame(
  Timestamp = ex_data$Timestamp,
  Axis1 = sample(0:200, 3200, TRUE, exp(rnorm(201, log(30), log(5)))),
  Axis2 = 0,
  Axis3 = 0,
  valid_status = sample(
    c("Awake-Wear", "Non-Wear", "Sleep"),
    3200,
    TRUE,
    c(2, 8, 14)
  )
)

full_data <- merge(ex_data, agd_data)

ee_data <- accelEE::ee_summary(full_data, scheme = "Hibbing 2023")

knitr::kable(ee_data)

## First, let's get our body composition data. We'll pretend it's the first row
## of the `info` data frame we created earlier. We'll remove the `ee` column to
## avoid confusion, and we'll also assign an ID label of `001` and a body mass
## of 87 kg

  body_comp <- info[1, ]
  body_comp$ee <- NULL
  body_comp$wt_kg <- 87

  
## Next, let's determine a Schofield prediction of basal metabolism to impute
## during periods of sleep/non-wear

  body_comp$schofield <- PAutilities::get_ree(
    method = "schofield_wt_ht", sex = "M", age_yr = 55,
    wt_kg = 87, ht_m = 1.85, output = "kcal_day"
  )


## It's time to merge the body composition and energy expenditure data

  ei_data <- merge(body_comp, ee_data)
  

## Next, we want to take our EE estimates (kcal/kg for awake wear periods) and
## convert them to total EE (i.e., no longer relative to body weight, and
## imputing basal metabolic expenditure for periods of sleep/non-wear).
  
## To do this, we'll use some functions from the `dplyr` package. Note that you
## can interpret the `%>%` operator as meaning 'next do ____' or 'then do ___'.
  
## The `%<>%` operator is for assignment (i.e., saves output to the `ei_data` object).
  
  suppressPackageStartupMessages(library(dplyr))
  
  ei_data %<>%
    mutate(across(
      total_hildebrand_linear:total_staudenmayer_rf,
      ~ .x * wt_kg + schofield * ((total_is_NonWear + total_is_Sleep) / 1440),
      .names = "{gsub(\"^total\", \"EE\", .col)}"
    )) %>%
    select(
      any_of(names(body_comp)),
      !c(Date, total_hildebrand_linear:total_staudenmayer_rf)
    ) %>%
    group_by(across(any_of(names(body_comp)))) %>%
    summarise(across(
      everything(), mean, .names = "{gsub(\"^total\", \"mean\", .col)}"
    ), .groups = "drop")

  
  knitr::kable(ei_data)

  ## Note that the `n_days` column refers to the number of days between
  ## body composition measurements, not necessarily the number of days of
  ## accelerometer data included in the analysis (n = 2 in this case)
  

## Generate the result
ei_result <-
  ei_data %>%
  mutate(across(
    EE_hildebrand_linear:EE_staudenmayer_rf,
    ~ IntakeBalance::IntakeBalance(
      fm_start, fm_end, ffm_start, ffm_end, ee_per_day = .x
    ),
    .names = "{gsub(\"^EE\", \"EI\", .col)}"
  ))

## Show the result
knitr::kable(ei_result)