Test for a change at a date fixed in advance

The question behind most applied changepoint work: the policy took effect on 1 March, the plant was retooled in week 14, the drug was approved in Q2. Did anything change then? Because the date was chosen before looking at the data, this is an ordinary two-sample comparison at a fixed split and needs no adjustment for selection: the p-value means what it says.

Usage

cpt_test_at(
  x,
  when,
  window = 0,
  change_in = c("mean", "var", "meanvar", "distribution"),
  family = c("gaussian", "poisson", "binomial", "exponential", "l1"),
  span = NULL,
  index = NULL,
  data = NULL,
  level = 0.95,
  B = 999,
  seed = NULL
)

# S3 method for class 'ggcpt_test_at'
print(x, ...)

Arguments

x: A numeric series (a vector, ts, zoo and so on), a ggcpt fit (its series and index are used), or a formula with data for a break in a regression (a Chow test).
when: When the change took effect: the first observation of the new regime, as a position or a value of the series' index (a date; for a numeric index such as a ts's years, a number outside 1..n is read as an index value). A detected changepoint cp (the last observation of the old regime) corresponds to when = cp + 1.
window: Allow the change anywhere within window observations either side of when. The statistic is then the largest over the window and its p-value comes from a permutation distribution of that maximum, which pays for the search. Defaults to 0: the date is exact.
change_in: What to test: "mean" (the default), "var", "meanvar" or "distribution".
family: For change_in = "mean": "gaussian" (a Welch t-test, the default), "poisson" (an exact test of two rates), "binomial" (Fisher's exact test of two proportions), "exponential" (an exact F test of two rates) or "l1" (a Wilcoxon rank-sum test, robust to outliers).
span: Optional number of observations each side of the split to compare. Defaults to all of them; a span keeps a change elsewhere in a long series out of the comparison.
index: Optional time index when x is a bare series.
data: A data frame, for formula input.
level: Confidence level for the interval on the change.
B: Permutations for window > 0. Defaults to 999.
seed: Optional seed for the permutations, scoped to this call.
...: Ignored.

Value

A ggcpt_test_at tibble with one row: when, cp (the last observation before it, in the package's convention) and cp_index, window, estimate (the change: a difference in means, a ratio of rates or variances), conf_low, conf_high, statistic, p_value, n_before, n_after, method and selection_adjusted (TRUE: the location was not chosen from the data, and a window's search is paid for by the permutation).

Examples

set.seed(1)
dates <- as.Date("2026-01-01") + 0:119
x <- c(rnorm(60), rnorm(60, 0.8))
cpt_test_at(x, when = as.Date("2026-03-02"), index = dates)
#> ggcpt_test_at (change in mean; the location was fixed in advance, so no selection adjustment is needed)
#> 
#> # A tibble: 1 × 13
#>   when          cp cp_index   window estimate conf_low conf_high statistic
#>   <date>     <int> <date>      <int>    <dbl>    <dbl>     <dbl>     <dbl>
#> 1 2026-03-02    60 2026-03-01      0    0.804    0.484      1.12      4.98
#> # ℹ 5 more variables: p_value <dbl>, n_before <int>, n_after <int>,
#> #   method <chr>, selection_adjusted <lgl>
# the policy took effect sometime in that fortnight
cpt_test_at(x, when = 61, window = 7, seed = 1)
#> ggcpt_test_at (change in mean; the location was fixed in advance, so no selection adjustment is needed)
#> 
#> # A tibble: 1 × 12
#>    when    cp window estimate conf_low conf_high statistic p_value n_before
#>   <dbl> <int>  <int>    <dbl>    <dbl>     <dbl>     <dbl>   <dbl>    <int>
#> 1    61    60      7    0.804    0.484      1.12      4.98   0.001       60
#> # ℹ 3 more variables: n_after <int>, method <chr>, selection_adjusted <lgl>

Usage

Arguments

Value

See also

Examples