R/nulldistswip.R

In hamsamilton/cgms.analysis: Functions for CGMS analysis

#by sampling differences and constructing a faux timeseries with them
addnulldistdiff <- function(df) {

  #initlist
  difflist <- c()
  for(i in unique(df$run_id)){
    filtereddf <- df %>% dplyr::filter(df$run_id == i)
    diffs <- diff(filtereddf$historic_glucose)
    difflist <- append(difflist,diffs)
  }

  #init nulldistframe
  nulldistframe <- data.frame(matrix(ncol = 21,nrow = length(unique(df$time_past_start))))
  names(nulldistframe) <- names(df)
  nulldistframe <- nulldistframe %>%
    dplyr::mutate_at(.vars = c("projno","subjno","visit","id","Subject",
                               "Age","Gender","Group","subj_id","run_id"),
                     function(x){
                       return("nulldist")
                     }
    )
  nulldistframe$time_past_start <- unique(df$time_past_start)
  difflist.resampled <- sample(difflist,size = length(unique(df$time_past_start)))

  nulldisttrace = c(mean(df$historic_glucose))
  for (i in 2:length(difflist.resampled)) {
    nulldisttrace[i] <- nulldisttrace[i-1] + difflist.resampled[i]#add current diff to previous value in nulldifftrance
  }
  nulldistframe$historic_glucose <- nulldisttrace
  rdf <- rbind(df,nulldistframe)
  return(rdf)
}

# As above, but with "Guard rails" for max and min to keep it constrained
addnulldistdiffguard <- function(df) {

  #initlist
  difflist <- c()
  for(i in unique(df$run_id)){
    filtereddf <- df %>% dplyr::filter(df$run_id == i)
    diffs <- diff(filtereddf$historic_glucose)
    difflist <- append(difflist,diffs)
  }

  #init nulldistframe
  nulldistframe <- data.frame(matrix(ncol = 21,nrow = length(unique(df$time_past_start))))
  names(nulldistframe) <- names(df)
  nulldistframe <- nulldistframe %>%
    dplyr::mutate_at(.vars = c("projno","subjno","visit","id","Subject",
                               "Age","Gender","Group","subj_id","run_id"),
                     function(x){
                       return("nulldist")
                     }
    )
  nulldistframe$time_past_start <- unique(df$time_past_start)

  maxbp <- max(df$historic_glucose)
  minbp <- min(df$historic_glucose)

  # init difflist
  difflist.resampled <- c(mean(df$historic_glucose))

  while(length(difflist.resampled) < nrow(nulldistframe)){
    sample.1 <- sample(difflist,size = 1)
    if(last(difflist.resampled) + sample.1 < maxbp & last(difflist.resampled) + sample.1 > minbp){
      difflist.resampled <- append(difflist.resampled,sample.1 + last(difflist.resampled))
    }
  }
  nulldisttrace = c(mean(df$historic_glucose))
  for (i in 2:length(difflist.resampled)) {
    nulldisttrace[i] <- nulldisttrace[i-1] + difflist.resampled[i]#add current diff to previous value in nulldifftrance
  }
  nulldistframe$historic_glucose <- difflist.resampled
  rdf <- rbind(df,nulldistframe)
  return(rdf)
}