tabler_by: tabler_by

In raredd/rawr: rawr miscellaneous

tabler_by

Description

This function is helpful for making simple, formatted tables similar to the functionality of tabular.

tabler_by creates simple tables, and tabler_by2 is a wrapper which can create stratified tables (tabler_by can also achieve this but requires additional steps).

Usage

tabler_by(
  data,
  varname,
  byvar,
  n = NULL,
  order = FALSE,
  zeros = TRUE,
  pct = FALSE,
  pct.column = FALSE,
  pct.total = FALSE,
  pct.sign = FALSE,
  drop = TRUE
)

tabler_by2(
  data,
  varname,
  byvar,
  n = NULL,
  order = FALSE,
  stratvar = NULL,
  zeros = TRUE,
  pct = FALSE,
  pct.column = FALSE,
  pct.total = FALSE,
  pct.sign = TRUE,
  drop = TRUE,
  collapse_varname = FALSE,
  collapse_format = c("<b>%s</b>", "%s")
)

Arguments

data	a data frame; variables varname and byvar should be factors
varname	subgroup variable name (rows)
byvar	stratification variable name (columns)
n	number in each group; see details
order	logical; order the result by decreasing frequency
zeros	optional character string replacement for cells which have zero counts; will appear as 0 (0%) if TRUE
pct	logical; if TRUE (and n is not missing), percents are shown
pct.column	logical; if TRUE, percents are separated into new columns
pct.total	logical; if TRUE, adds percents for total column
pct.sign	logical; if TRUE, percent sign is shown; otherwise, percents are shown in parens without sign
drop	logical; for tabler_by if TRUE, rows or columns with zero total counts will be removed (default); the FALSE case is useful when merging multiple tabler_by tables (e.g., this is how tabler_by2 aligns stratified tables)
stratvar	for tabler_by2, a factor-like variable used to stratify observations into mutually exclusive groups for which tabler_by will be performed on each subset
collapse_varname	logical; for tabler_by2, combine multiple varnames into a single column with indents and extra rows for headers
collapse_format	format strings for headers and normal rows, respectively, passed to sprintf

Details

varname and byvar should be factors, and the levels will appear in the output as they occur in levels(x).

n is used to calculate the percent. If missing, the output will only show counts in the table. If given, length(n) should be one or equal to the number of levels of byvar.

If one n is given, tabler_by assumes that this is the total population for a subgroup, e.g., if creating a table for a subset of the data, it is only necessary to provide the total n for that group.

If more than one n is given, tabler_by assumes that the entire data set is given as data and will use the corresponding n for percents.

See Also

tox_worst; match_ctc

Other tabler: tabler_resp(), tabler_stat2(), tabler_stat(), tabler()

Examples

mt <- within(mtcars, {
  am   <- factor(am)
  gear <- factor(gear)
  vs   <- factor(vs, 0:2)
  vs2  <- factor(vs, 2:0)
  carb <- factor(carb)
})

tabler_by(mt, 'vs', 'gear')
tabler_by(mt, c('vs2', 'carb'), 'gear', order = FALSE)
tabler_by(mt, c('vs2', 'carb'), 'gear', order = TRUE)


## when length(n) > 1, each column uses a different n for percents
tabler_by(mt, 'vs', 'gear', n = 32, pct = TRUE)
tabler_by(mt, 'vs', 'gear', n = table(mt$gear), pct = TRUE)
tabler_by(mt, 'vs', 'gear', n = table(mt$gear), zeros = '-',
          pct = TRUE, pct.column = TRUE, pct.total = TRUE)


## use tabler_by2 to create a stratified table
t1 <- tabler_by2(mt, c('vs', 'carb'), 'gear', stratvar = 'am', order = TRUE)

## or tabler_by to do this in several steps
t2 <- cbind(
  tabler_by(mt, varname = c('vs', 'carb'), byvar = 'gear',
            drop = FALSE)[, c('carb', 'Total')],
  tabler_by(mt[mt$am == 0, ], varname = c('vs', 'carb'), byvar = 'gear',
            drop = FALSE)[, -1],
  tabler_by(mt[mt$am == 1, ], varname = c('vs', 'carb'), byvar = 'gear',
            drop = FALSE)[, -1]
)

## order, drop extra rows/columns, set rownames
rownames(t2) <- locf(rownames(t2))
t2 <- t2[order(locf(rownames(t2)), -xtfrm(t2[, 2])), ]
t2 <- t2[!t2[, 'Total'] %in% '0', ]
t2 <- t2[, apply(t2, 2, function(x) !all(x %in% '0'))]
rownames(t2)[duplicated(rownames(t2))] <- ''

stopifnot(identical(t1, t2))


## example workflow
set.seed(1)
f <- function(x, ...) sample(x, 100, replace = TRUE, ...)
tox <- data.frame(
  id = rep(1:10, 10), phase = 1:2,
  code = f(rawr::ctcae_v4$tox_code[1:100]),
  grade = f(1:3, prob = c(0.6, 0.3, 0.1))
)

tox <- cbind(tox, match_ctc(tox$code)[, c('tox_cat', 'tox_desc')])

## get worst toxicities by id, by grade
n <- colSums(table(tox$id, tox$phase) > 0)
tox[] <- lapply(tox, factor)
tox <- tox_worst(tox, desc = 'tox_desc')

out <- tabler_by2(tox, 'tox_desc', 'grade', stratvar = 'phase', zeros = '-')
colnames(out)[1] <- sprintf('Total<br /><font size=1>n = %s</font>', sum(n))
cgroup <- c(
  '',
  sprintf('Phase I<br /><font size=1>n = %s</font>', n[1]),
  sprintf('Phase II<br /><font size=1>n = %s</font>', n[2])
)

htmlTable::htmlTable(
  out, ctable = TRUE, cgroup = cgroup, n.cgroup = c(1, 4, 4),
  caption = 'Table 1: Toxicities<sup>&dagger;</sup> by phase and grade,
            sorted by total.',
  col.columns = rep(c('grey97', 'none', 'grey97'), times = c(1, 4, 4)),
  col.rgroup = rep(rep(c('none', 'grey97'), each = 5), 10),
  tfoot = paste0('<font size=1><sup>&dagger;</sup>Percents represent ',
           'proportion of patients out of respective phase total.</font>')
)


## same as above but add level of stratification, sort by total within group
out2 <- tabler_by2(tox, c('tox_cat', 'tox_desc'), 'grade', order = TRUE,
                   stratvar = 'phase', zeros = '-', n = c(5, 5), pct = TRUE)
stopifnot(
  identical(
    sort(unname(out[, grep('Total', colnames(out))[1]])),
    sort(unname(out2[, grep('Total', colnames(out2))[1]]))
))

colnames(out2)[1:2] <- c(
  'Description', sprintf('Total<br /><font size=1>n = %s</font>', sum(n))
)

cgroup <- c(
  '', '',
  sprintf('Phase I<br /><font size=1>n = %s</font>', n[1]),
  sprintf('Phase II<br /><font size=1>n = %s</font>', n[2])
)

htmlTable::htmlTable(
  out2, align = 'lc', cgroup = cgroup, n.cgroup = c(1, 1, 4, 4),
  caption = 'Table 1: Toxicities<sup>&dagger;</sup> by category, phase,
  grade.'
)


## collapse varname columns into single column with indents/extra rows
out3 <- tabler_by2(
  tox, c('tox_cat', 'tox_desc'), 'grade', stratvar = 'phase', zeros = '-',
  n = c(5, 5), pct = TRUE, pct.sign = FALSE,
  collapse_varname = TRUE, collapse_format = c('<i>%s</i>', '&emsp;%s')
)

htmlTable::htmlTable(
  out3, align = 'c', cgroup = cgroup[-1], n.cgroup = c(1, 1, 4, 4)[-1],
  caption = 'Table 1: Toxicities<sup>&dagger;</sup> by category, phase,
  grade.'
)

raredd/rawr documentation built on March 4, 2024, 1:36 a.m.