Convert p-values to (pseudo) Bayes Factors. This transformation has been
suggested by Wagenmakers (2022), but is based on a vast amount of assumptions.
It might therefore be not reliable. Use at your own risks. For more accurate
approximate Bayes factors, use bic_to_bf()
instead.
Usage
p_to_bf(x, log = FALSE, ...)
# S3 method for numeric
p_to_bf(x, log = FALSE, n_obs = NULL, ...)
# S3 method for default
p_to_bf(x, log = FALSE, ...)
Arguments
- x
A (frequentist) model object, or a (numeric) vector of p-values.
- log
Wether to return log Bayes Factors. Note: The
print()
method always showsBF
- the"log_BF"
column is only accessible from the returned data frame.- ...
Other arguments to be passed (not used for now).
- n_obs
Number of observations. Either length 1, or same length as
p
.
References
Wagenmakers, E.J. (2022). Approximate objective Bayes factors from p-values and sample size: The 3p(sqrt(n)) rule. Preprint available on ArXiv: https://psyarxiv.com/egydq
See also
bic_to_bf()
for more accurate approximate Bayes factors.
Examples
if (requireNamespace("parameters", quietly = TRUE)) {
data(iris)
model <- lm(Petal.Length ~ Sepal.Length + Species, data = iris)
p_to_bf(model)
# Examples that demonstrate comparison between
# BIC-approximated and pseudo BF
# --------------------------------------------
m0 <- lm(mpg ~ 1, mtcars)
m1 <- lm(mpg ~ am, mtcars)
m2 <- lm(mpg ~ factor(cyl), mtcars)
# In this first example, BIC-approximated BF and
# pseudo-BF based on p-values are close...
# BIC-approximated BF, m1 against null model
bic_to_bf(BIC(m1), denominator = BIC(m0))
# pseudo-BF based on p-values - dropping intercept
p_to_bf(m1)[-1, ]
# The second example shows that results from pseudo-BF are less accurate
# and should be handled wit caution!
bic_to_bf(BIC(m2), denominator = BIC(m0))
p_to_bf(anova(m2), n_obs = nrow(mtcars))
}
#> Pseudo-BF (against NULL)
#>
#> Parameter | p | BF
#> -------------------------------
#> factor(cyl) | < .001 | 1.18e+07