AMR/R/freq.R

362 lines
14 KiB
R

# ==================================================================== #
# TITLE #
# Antimicrobial Resistance (AMR) Analysis #
# #
# AUTHORS #
# Berends MS (m.s.berends@umcg.nl), Luz CF (c.f.luz@umcg.nl) #
# #
# LICENCE #
# This program is free software; you can redistribute it and/or modify #
# it under the terms of the GNU General Public License version 2.0, #
# as published by the Free Software Foundation. #
# #
# This program is distributed in the hope that it will be useful, #
# but WITHOUT ANY WARRANTY; without even the implied warranty of #
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the #
# GNU General Public License for more details. #
# ==================================================================== #
#' Frequency table
#'
#' Create a frequency table of a vector of data, a single column or a maximum of 9 columns of a data frame. Supports markdown for reports.
#' @param x data
#' @param sort.count Sort on count. Use \code{FALSE} to sort alphabetically on item.
#' @param nmax number of row to print. Use \code{nmax = 0} or \code{nmax = NA} to print all rows.
#' @param na.rm a logical value indicating whether NA values should be removed from the frequency table. The header will always print the amount of\code{NA}s.
#' @param markdown print table in markdown format (this forces \code{nmax = NA})
#' @param toConsole Print table to the console. Use \code{FALSE} to assign the table to an object.
#' @param digits how many significant digits are to be used for numeric values (not for the items themselves, that depends on \code{\link{getOption}("digits")})
#' @param sep a character string to separate the terms when selecting multiple columns
#' @details For numeric values, the next values will be calculated and shown into the header:
#' \itemize{
#' \item{Mean, using \code{\link[base]{mean}}}
#' \item{Standard deviation, using \code{\link[stats]{sd}}}
#' \item{Five numbers of Tukey (min, Q1, median, Q3, max), using \code{\link[stats]{fivenum}}}
#' \item{Outliers (count and list), using \code{\link{boxplot.stats}}}
#' \item{Coefficient of variation (CV), the standard deviation divided by the mean}
#' \item{Coefficient of quartile variation (CQV, sometimes called coefficient of dispersion), calculated as \code{(Q3 - Q1) / (Q3 + Q1)} using \code{\link{quantile}} with \code{type = 6} as quantile algorithm to comply with SPSS standards}
#' }
#' @importFrom stats fivenum sd quantile
#' @importFrom grDevices boxplot.stats
#' @importFrom dplyr %>% select pull n_distinct group_by arrange desc mutate summarise
#' @keywords summary summarise frequency freq
#' @rdname freq
#' @export
#' @examples
#' library(dplyr)
#'
#' freq(septic_patients$hospital_id)
#'
#' septic_patients %>%
#' filter(hospital_id == "A") %>%
#' select(bactid) %>%
#' freq()
#'
#' # select multiple columns; they will be pasted together
#' septic_patients %>%
#' left_join_microorganisms %>%
#' filter(hospital_id == "A") %>%
#' select(genus, species) %>%
#' freq()
#'
#' # save frequency table to an object
#' years <- septic_patients %>%
#' mutate(year = format(date, "%Y")) %>%
#' select(year) %>%
#' freq(toConsole = FALSE)
freq <- function(x,
sort.count = TRUE,
nmax = 15,
na.rm = TRUE,
markdown = FALSE,
toConsole = TRUE,
digits = 2,
sep = " ") {
mult.columns <- 0
if (NROW(x) == 0) {
cat('\nNo observations.\n')
return(invisible())
}
if (!is.null(ncol(x))) {
if (ncol(x) == 1 & any(class(x) == 'data.frame')) {
x <- x %>% pull(1)
} else if (ncol(x) < 10) {
mult.columns <- ncol(x)
colnames(x) <- LETTERS[1:ncol(x)]
if (ncol(x) == 2) {
x$total <- paste(x$A %>% as.character(),
x$B %>% as.character(),
sep = sep)
} else if (ncol(x) == 3) {
x$total <- paste(x$A %>% as.character(),
x$B %>% as.character(),
x$C %>% as.character(),
sep = sep)
} else if (ncol(x) == 4) {
x$total <- paste(x$A %>% as.character(),
x$B %>% as.character(),
x$C %>% as.character(),
x$D %>% as.character(),
sep = sep)
} else if (ncol(x) == 5) {
x$total <- paste(x$A %>% as.character(),
x$B %>% as.character(),
x$C %>% as.character(),
x$D %>% as.character(),
x$E %>% as.character(),
sep = sep)
} else if (ncol(x) == 6) {
x$total <- paste(x$A %>% as.character(),
x$B %>% as.character(),
x$C %>% as.character(),
x$D %>% as.character(),
x$E %>% as.character(),
x$F %>% as.character(),
sep = sep)
} else if (ncol(x) == 7) {
x$total <- paste(x$A %>% as.character(),
x$B %>% as.character(),
x$C %>% as.character(),
x$D %>% as.character(),
x$E %>% as.character(),
x$F %>% as.character(),
x$G %>% as.character(),
sep = sep)
} else if (ncol(x) == 8) {
x$total <- paste(x$A %>% as.character(),
x$B %>% as.character(),
x$C %>% as.character(),
x$D %>% as.character(),
x$E %>% as.character(),
x$F %>% as.character(),
x$G %>% as.character(),
x$H %>% as.character(),
sep = sep)
} else if (ncol(x) == 9) {
x$total <- paste(x$A %>% as.character(),
x$B %>% as.character(),
x$C %>% as.character(),
x$D %>% as.character(),
x$E %>% as.character(),
x$F %>% as.character(),
x$G %>% as.character(),
x$H %>% as.character(),
x$I %>% as.character(),
sep = sep)
}
x <- x$total
} else {
stop('A maximum of 9 columns can be analysed at the same time.', call. = FALSE)
}
}
if (markdown == TRUE & toConsole == FALSE) {
warning('`toConsole = FALSE` will be ignored when `markdown = TRUE`.')
}
if (mult.columns > 1) {
NAs <- x[is.na(x) | x == trimws(strrep('NA ', mult.columns))]
} else {
NAs <- x[is.na(x)]
}
if (na.rm == TRUE) {
x <- x[!x %in% NAs]
}
if (missing(sort.count) & any(class(x) %in% c('double', 'integer', 'numeric', 'raw', 'single', 'factor'))) {
# sort on item/level at default when x is numeric or a factor and sort.count is not set
sort.count <- FALSE
}
header <- character(0)
markdown_line <- ''
if (markdown == TRUE) {
markdown_line <- '\n'
}
x_align <- 'l'
if (mult.columns > 0) {
header <- header %>% paste0(markdown_line, 'Columns: ', mult.columns)
} else {
header <- header %>% paste0(markdown_line, 'Class: ', class(x) %>% rev() %>% paste(collapse = " > "))
}
if (is.list(x) | is.matrix(x) | is.environment(x) | is.function(x)) {
cat(header, "\n")
stop('`freq()` does not support lists, matrices, environments or functions.', call. = FALSE)
}
header <- header %>% paste0(markdown_line, '\nLength: ', (NAs %>% length() + x %>% length()) %>% format(),
' (of which NA: ', NAs %>% length() %>% format(),
' = ', (NAs %>% length() / (NAs %>% length() + x %>% length())) %>% percent(force_zero = TRUE), ')')
header <- header %>% paste0(markdown_line, '\nUnique: ', x %>% n_distinct() %>% format())
header.numbers.done <- FALSE
if (any(class(x) %in% c('double', 'integer', 'numeric', 'raw', 'single'))) {
# right align number
x_align <- 'r'
header <- header %>% paste0('\n')
header <- header %>% paste(markdown_line, '\nMean: ', x %>% base::mean(na.rm = TRUE) %>% format(digits = digits))
header <- header %>% paste0(markdown_line, '\nStd. dev.: ', x %>% stats::sd(na.rm = TRUE) %>% format(digits = digits),
' (CV: ', x %>% cv(na.rm = TRUE) %>% format(digits = digits), ')')
header <- header %>% paste0(markdown_line, '\nFive-Num: ', x %>% stats::fivenum(na.rm = TRUE) %>% format(digits = digits) %>% trimws() %>% paste(collapse = ' | '),
' (CQV: ', x %>% cqv(na.rm = TRUE) %>% format(digits = digits), ')')
outlier_length <- length(boxplot.stats(x)$out)
header <- header %>% paste0(markdown_line, '\nOutliers: ', outlier_length)
if (outlier_length > 0) {
header <- header %>% paste0(' (unique: ', boxplot.stats(x)$out %>% unique() %>% length(), ')')
}
}
formatdates <- "%e %B %Y" # = d mmmm yyyy
if (any(class(x) == 'hms')) {
x <- x %>% as.POSIXlt()
formatdates <- "%H:%M:%S"
}
if (any(class(x) %in% c('Date', 'POSIXct', 'POSIXlt'))) {
header <- header %>% paste0('\n')
mindatum <- x %>% min()
maxdatum <- x %>% max()
header <- header %>% paste0(markdown_line, '\nOldest: ', mindatum %>% format(formatdates) %>% trimws())
header <- header %>% paste0(markdown_line, '\nNewest: ', maxdatum %>% format(formatdates) %>% trimws(),
' (+', difftime(maxdatum, mindatum, units = 'auto') %>% as.double() %>% format(), ')')
}
if (any(class(x) == 'POSIXlt')) {
x <- x %>% format(formatdates)
}
if (toConsole == TRUE) {
cat(header)
}
if (all(is.na(x))) {
cat('\n\nNo observations.\n')
return(invisible())
}
if (n_distinct(x) == length(x)) {
warning('All observations are unique.', call. = FALSE)
}
if (nmax == 0 | is.na(nmax)) {
nmax <- length(x)
}
nmax.1 <- min(length(x), nmax + 1)
# create table with counts and percentages
if (any(class(x) == 'factor')) {
df <- tibble::tibble(Item = x,
Fctlvl = x %>% as.integer()) %>%
group_by(Item, Fctlvl)
column_names <- c('Item', 'Count', 'Percent', 'Cum. Count', 'Cum. Percent', '(Factor Level)')
column_align <- c('l', 'r', 'r', 'r', 'r', 'r')
} else {
df <- tibble::tibble(Item = x) %>%
group_by(Item)
column_names <- c('Item', 'Count', 'Percent', 'Cum. Count', 'Cum. Percent')
column_align <- c(x_align, 'r', 'r', 'r', 'r')
}
df <- df %>%
summarise(Count = n(),
Percent = (n() / length(x)) %>% percent(force_zero = TRUE))
if (df$Item %>% paste(collapse = ',') %like% '\033') {
df <- df %>%
mutate(Item = Item %>%
# remove escape char
# see https://en.wikipedia.org/wiki/Escape_character#ASCII_escape_character
gsub('\033', ' ', ., fixed = TRUE))
}
# sort according to setting
if (sort.count == TRUE) {
df <- df %>% arrange(desc(Count))
} else {
if (any(class(x) == 'factor')) {
df <- df %>% arrange(Fctlvl)
} else {
df <- df %>% arrange(Item)
}
}
# add cumulative values
df$Cum <- cumsum(df$Count)
df$CumTot <- (df$Cum / sum(df$Count, na.rm = TRUE)) %>% percent(force_zero = TRUE)
df$Cum <- df$Cum %>% format()
if (any(class(x) == 'factor')) {
# put factor last
df <- df %>% select(Item, Count, Percent, Cum, CumTot, Fctlvl)
}
if (markdown == TRUE) {
tblformat <- 'markdown'
} else {
tblformat <- 'pandoc'
}
if (toConsole == FALSE) {
# assign to object
df[, 3] <- df[, 2] / sum(df[, 2], na.rm = TRUE)
df[, 4] <- cumsum(df[, 2])
df[, 5] <- df[, 4] / sum(df[, 2], na.rm = TRUE)
return(df)
} else {
# save old NA setting for kable
opt.old <- options()$knitr.kable.NA
options(knitr.kable.NA = "<NA>")
Count.rest <- sum(df[nmax.1:nrow(df), 'Count'], na.rm = TRUE)
if (any(class(x) %in% c('double', 'integer', 'numeric', 'raw', 'single'))) {
df <- df %>% mutate(Item = format(Item))
}
df <- df %>% mutate(Count = format(Count))
if (nrow(df) > nmax.1 & markdown == FALSE) {
df2 <- df[1:nmax,]
print(
knitr::kable(df2,
format = tblformat,
col.names = column_names,
align = column_align,
padding = 1)
)
cat('... and ',
format(nrow(df) - nmax),
' more ',
paste0('(n = ',
format(Count.rest),
'; ',
(Count.rest / length(x)) %>% percent(force_zero = TRUE),
')'),
'. Use `nmax` to show more rows.\n', sep = '')
} else {
print(
knitr::kable(df,
format = tblformat,
col.names = column_names,
align = column_align,
padding = 1)
)
}
cat('\n')
# reset old kable setting
options(knitr.kable.NA = opt.old)
return(invisible())
}
}
#' @rdname freq
#' @export
frequency_tbl <- freq