Computes a Bayes factor quantifying evidence that a reported p-value is inconsistent with the reported test statistic.
Usage
bayes_verify_pvalue(
test_type = c("t", "z", "chi_sq", "f"),
statistic,
df = NULL,
reported_p,
alternative = "two.sided",
rounding_sd = 0.005,
fabrication_sd = 0.15
)Arguments
- test_type
Character. One of
"t","z","chi_sq","f".- statistic
Numeric. The reported test statistic.
- df
Numeric. Degrees of freedom. For F-tests, a vector of length 2. Not required for z-tests.
- reported_p
Numeric. The reported p-value.
- alternative
Character. One of
"two.sided"(default),"less","greater".- rounding_sd
Numeric. SD of the rounding error model (default 0.005).
- fabrication_sd
Numeric. SD under the fabrication model (default 0.15).
Value
A list with components:
- bf_inconsistent
Numeric. Bayes factor for inconsistency.
- posterior_prob_inconsistent
Numeric. Posterior probability.
- recalculated_p
Numeric. Recalculated p-value.
- discrepancy
Numeric. Absolute difference.
- interpretation
Character. Evidence strength label.
Details
Discrepancy modelled as N(0, rounding_sd^2) under H0 (honest rounding) and N(0, fabrication_sd^2) under H1 (fabrication/error).
Examples
bayes_verify_pvalue("chi_sq", statistic = 3.84, df = 1, reported_p = 0.05)
#> $bf_inconsistent
#> [1] 0.03333459
#>
#> $posterior_prob_inconsistent
#> [1] 0.03225925
#>
#> $recalculated_p
#> [1] 0.05004352
#>
#> $discrepancy
#> [1] 4.352125e-05
#>
#> $interpretation
#> [1] "consistent"
#>
#> $test_type
#> [1] "chi_sq"
#>
#> $statistic
#> [1] 3.84
#>