When a detection comes back empty, the useful question is whether a
change of the size you care about would have been found. This runs
cpt_min_detectable() at the fit's own length, method and
penalty, with the noise level estimated from the series, and reports the
smallest shift detected with the requested power: changes smaller than
that could well be there and missed.
Usage
cpt_null_power(fit, power = 0.8, n_sim = 50, seed = NULL, ...)
# S3 method for class 'ggcpt_null_power'
print(x, ...)Arguments
- fit
A
ggcptobject (usually one with no changepoints; it works for any).- power
Target power. Defaults to
0.8.- n_sim
Simulations per power evaluation. Defaults to
50.- seed
Optional seed, scoped to this call.
- ...
Further arguments for
cpt_min_detectable().- x
A
ggcpt_null_powerobject.
Value
A ggcpt_null_power list with n, sigma (the
noise level used), jump (the smallest detectable shift, in the
data's units), jump_sd (the same in noise standard deviations),
power, method and constant (TRUE when the
series has no variation, in which case nothing is simulated). With a
print() method.
See also
Other inference:
cpt_assumptions(),
cpt_attribute_event(),
cpt_effect(),
cpt_gof(),
cpt_robustness(),
cpt_test_at(),
cpt_test_null()
Examples
# \donttest{
set.seed(1)
fit <- cpt_detect(rnorm(100), method = "pelt")
cpt_null_power(fit, n_sim = 20, seed = 1)
#> ggcpt_null_power (method: pelt, n = 100)
#> At this length and noise level (sd 0.961), a single shift of about 1.09 (1.13 sd) is
#> detected with power 0.8. A smaller change could be present and missed.
# }