table1() computes the numbers of a descriptive table, per group and for
all rows. It does no formatting.
Value
A data.table with one row per variable and group, in the order of
vars, and "All" first. The columns are var, grp, n, missing,
mean, sd, events, pct, p and test. n counts the non-missing
values.
Details
The groups are the levels of by, plus "All". The levels are the factor
levels when by is a factor, and the sorted unique values otherwise. A row
with a missing by counts in "All" only. A level of by that is "All"
is an error. A level with no rows has a row in the output with n = 0.
For a "bin" variable:
eventsis the count of 1, andpctis100 * events / n.pcompares the levels ofbythat have a non-missing value. It comes from the Pearson chi-square test without continuity correction. When any expected count is below 5, it comes from Fisher's exact test.testis"chisq"or"fisher", and it is the same on every row of the variable. When fewer than two levels have a non-missing value,pandtestareNA.A value other than 0, 1 or
NAis an error.
For a "cont" variable, mean and sd hold the numbers. p and test
are NA in this version.
A column that does not apply to the type holds NA.
Examples
d <- data.table::data.table(
g = rep(c("a", "b"), each = 10),
smoker = rep(c(1, 0, 0, 1), 5),
age = 21:40
)
vars <- data.table::data.table(
var = c("smoker", "age"),
type = c("bin", "cont")
)
pbn::table1(d, by = "g", vars = vars)
#> var grp n missing mean sd events pct p test
#> <char> <char> <int> <int> <num> <num> <int> <num> <num> <char>
#> 1: smoker All 20 0 NA NA 10 50 1 chisq
#> 2: smoker a 10 0 NA NA 5 50 1 chisq
#> 3: smoker b 10 0 NA NA 5 50 1 chisq
#> 4: age All 20 0 30.5 5.91608 NA NA NA <NA>
#> 5: age a 10 0 25.5 3.02765 NA NA NA <NA>
#> 6: age b 10 0 35.5 3.02765 NA NA NA <NA>
