Skip to contents

Tidy a Mechanism-Consistency Result

Source: R/tidy.R
tidy.mechanism_consistency_result.Rd

Turns a mechanism_consistency_test() verdict into a flat, broom-style one-row summary. Registered against the tidy() generic (re-exported by broom), so broom::tidy() and generics::tidy() both dispatch here.

Usage

# S3 method for class 'mechanism_consistency_result'
tidy(x, ...)

Arguments

x

A mechanism_consistency_result object from mechanism_consistency_test().

...

Currently unused; present for generic compatibility.

Value

A one-row data.frame with columns:

term

Character. Always "mechanism_consistency".

statistic

Numeric. The observed bd-HSIC statistic.

p.value

Numeric. The H0 tail p-value (p_h0).

decision

Character. The three-way decision enum ("no_effect", "mechanism_consistent_effect", "mechanism_inconsistent_effect").

grounding

Character. "grounded" when the mechanism's provenance was declared, else "ungrounded" (Independent Oracle Principle).

h1_percentile

Numeric. Where the observed statistic sits in the model-implied H1 band, in [0, 1].

borderline

Logical. Whether the consistency label is fragile.

n

Integer. The sample size.

Details

The p.value column carries the broom-canonical dot and reports the H0 tail p-value (p_h0). The grounding column threads through the Independent Oracle Principle label, so a downstream summary cannot silently drop the fact that a verdict built on an un-declared mechanism is "ungrounded".

See also

mechanism_consistency_test()

Examples

set.seed(1)
n <- 60L
nrate <- runif(n, 0, 200)
yield <- 1.1 + 4.2 * (1 - exp(-0.018 * nrate)) + rnorm(n, 0, 0.25)
post <- cbind(ymax = rnorm(150L, 4.2, 0.30),
              rate = rnorm(150L, 0.018, 0.004),
              y0   = rnorm(150L, 1.1, 0.15))
mitscherlich <- function(theta, X, t) {
  theta[3L] + theta[1L] * (1 - exp(-theta[2L] * t))
}
res <- mechanism_consistency_test(post, mitscherlich, X = matrix(1, n, 1L),
                 treatment = nrate, outcome = yield,
                 n_perm = 80L, seed = 1L)
tidy(res)
#>                    term  statistic    p.value                    decision
#> 1 mechanism_consistency 0.06587821 0.01234568 mechanism_consistent_effect
#>    grounding h1_percentile borderline  n
#> 1 ungrounded         0.425      FALSE 60