Skip to contents

Scan Cutpoints of Several Features with the Log-Rank Test

Source: R/batch_cutpoint_survdiff.R
batch_cutpoint_survdiff.Rd

For every feature, dichotomises the cohort at each candidate cutpoint (`<= cutpoint` is "low", `> cutpoint` "high"), runs [survival::survdiff()] and returns the p-values with a significance bin. Candidate cutpoints are the midpoints between consecutive distinct values or quantiles. The univariate Cox coefficient is returned too, which [first_significant_cutpoint()] uses to pick the first significant cutpoint.

Usage

batch_cutpoint_survdiff(
  data,
  time_col,
  status_col,
  feature_cols,
  cutpoints = c("midpoints", "quantiles"),
  probs = seq(0.1, 0.9, by = 0.1),
  min_n = 1,
  p_breaks = c(0.001, 0.05, 0.1)
)

Arguments

data

A data frame.

time_col, status_col

Names of the follow-up time and event (1 = event) columns.

feature_cols

Names of the numeric features to scan.

cutpoints

`"midpoints"` (default) or `"quantiles"`.

probs

Quantile probabilities for `cutpoints = "quantiles"`.

min_n

Minimum number of subjects in each of the two groups (default 1).

p_breaks

Increasing p-value breaks for the significance bin (default `c(0.001, 0.05, 0.1)`).

Value

A tibble with `feature`, `cutpoint`, `n_low`, `n_high`, `chisq`, `p`, `p_type` (ordered factor: `"p >= 0.1"`, `"p < 0.1"`, `"p < 0.05"`, `"p < 0.001"` for the default breaks), `cox_coef` and `cox_p`. A feature without a usable cutpoint contributes no rows.

See also

[first_significant_cutpoint()], [km_cutpoint_search()]

Examples

set.seed(1)
d <- data.frame(time = rexp(80), status = rbinom(80, 1, 0.7), x = rnorm(80), z = rnorm(80))
res <- batch_cutpoint_survdiff(d, "time", "status", c("x", "z"), cutpoints = "quantiles")
head(res)
#> # A tibble: 6 × 9
#>   feature cutpoint n_low n_high   chisq     p p_type   cox_coef cox_p
#>   <chr>      <dbl> <int>  <int>   <dbl> <dbl> <ord>       <dbl> <dbl>
#> 1 x        -1.11       8     72 1.02    0.313 p >= 0.1  0.00479 0.975
#> 2 x        -0.949     16     64 0.00354 0.953 p >= 0.1  0.00479 0.975
#> 3 x        -0.714     24     56 0.00167 0.967 p >= 0.1  0.00479 0.975
#> 4 x        -0.438     32     48 0.0322  0.858 p >= 0.1  0.00479 0.975
#> 5 x        -0.160     40     40 0.215   0.643 p >= 0.1  0.00479 0.975
#> 6 x         0.0963    48     32 0.0359  0.850 p >= 0.1  0.00479 0.975