General distribution functions for censored data

TDA's gdf: the marginal (or, with method, joint) distribution or survivor function for possibly right-censored data, via a Kaplan-Meier-style calculation – the same underlying censoring machinery as tda_lsreg1, but estimating the distribution function directly rather than a conditional expectation.

Usage

tda_gdf(
  formula,
  data,
  censor = NULL,
  id = NULL,
  dimension = NULL,
  what = c("distribution", "survivor", "expected_values"),
  method = c("marginal", "joint1", "joint2"),
  offset = NULL,
  n_boxes = NULL,
  delta = NULL,
  control = NULL,
  options = list(),
  dir = tempfile("tda"),
  ...
)

Arguments

formula: a one-sided formula naming the (possibly censored) variable, e.g. ~ Y, or a list of such formulas – one per dimension – for the multi-dimensional, grp=-based form (see tda_lsreg1's Details for how id groups rows into units).
data: a data frame.
censor: the censoring indicator, as a name or a vector, or NULL (the default) when every observation is exact: TRUE or 1 marks a case whose value is censored (the opposite of survival::Surv's event indicator). TDA's cen= is 1 for an exact observation and is translated here.
id: for multi-dimensional data, the name of the column identifying which unit each row belongs to (TDA's grp=).
dimension: with id, the name of the column saying which margin (dimension) each of a unit's observations belongs to – TDA's level variable in grp=ID,L1. Positive integers, not necessarily contiguous; TDA requires that no two observations of one unit share a value, so this is mandatory whenever an id value repeats. The data are in long format: one row per (unit, margin), and the joint methods estimate the joint distribution over the margins.
what: the distribution function (default), the survivor function, or the expected value of each case's observation under the fitted distribution – gdf's prn=. The third option returns a differently-shaped table: one row per case (not per distinct value), carrying the observed value, its exact/censored status (1 = exact, TDA's cen value), and its expected value.
method: "marginal" (default) or, for multi-dimensional data (with id), "joint1"/"joint2" – gdf's own opt=, the same three values tda_lsreg1's method takes.
offset, n_boxes: for method = "marginal" or "joint1" only, the domain offset and the number of grid boxes (gdf's sc=/n=).
delta: for method = "joint2" only, the delta grid spacing (gdf's d=).
control: for method = "marginal" or "joint1" only, convergence settings (only tda_control's maxit/tolf fields apply, as gdf's mxit=/tolf=).
options: a named list of further TDA options, passed through.
dir: working directory.
...: passed to tda_run.

Value

An object carrying a table of values and the distribution (or survivor) function at each.

Examples

set.seed(1)
n <- 20
d <- data.frame(y = round(rexp(n, 0.2), 1), cen = rbinom(n, 1, 0.3))
fit <- tda_gdf(~ y, d, censor = "cen")
fit$table
#>    dimension value distribution_function
#> 1          1   0.7             0.1000000
#> 2          1   1.7             0.1529412
#> 3          1   2.2             0.2058824
#> 4          1   2.7             0.2588235
#> 5          1   2.9             0.3117647
#> 6          1   3.3             0.3647059
#> 7          1   3.8             0.4176471
#> 8          1   5.2             0.4823529
#> 9          1   6.1             0.5686275
#> 10         1   6.2             0.6549020
#> 11         1   7.0             0.7411765
#> 12         1   9.4             0.8274510
#> 13         1  22.1             1.0000000

# expected values: one row per case, not per distinct value
fit2 <- tda_gdf(~ y, d, censor = "cen", what = "expected_values")
head(fit2$table)
#>   dimension    y exact  expected
#> 1         1  3.8     1  3.800000
#> 2         1  5.9     0 12.150000
#> 3         1  0.7     1  0.700000
#> 4         1  0.7     0  8.338562
#> 5         1  2.2     1  2.200000
#> 6         1 14.5     0 22.100000

General distribution functions for censored data

Usage

Arguments

Value

See also

Examples