RBF_ttest: Calculate a Replication Bayes Factor for a two-sample t-Test.

In neurotroph/ReplicationBF: Calculating Replication Bayes Factors for Different Scenarios

Description

RBF_ttest calculates a Replication Bayes Factor (Verhagen & Wagenmakers, 2014) for t-Tests.

Usage

RBF_ttest(t.orig, n.orig, t.rep, n.rep, method = "NormApprox",
  M = 1e+05, store.samples = FALSE)

Arguments

t.orig	t-statistic of the original study.
n.orig	Numeric vector (2 elements) with cell sizes.
t.rep	t-tstatistic of the replication study.
n.rep	Numeric vector (2 elements) with cell sizes.
method	Either "NormApprox" for a Normal approximation of the original study's posterior distribution or "MCMC" for an approximation through the Metropolis algorithm.
M	Number of posterior samples (either MCMC or Normal distribution).
store.samples	If TRUE, the returned object contains M samples from the original study's posterior distribution (required for plot_RBF).

Details

The Replication Bayes Factor is a marginal likelihood ratio between two models, characterized by the following positions:

• H0: The position of a skeptic, who does not place confidence in the findings of the original study and assumes θ \approx 0.

• Hr: The position of a proponent, who expects the original study to be a faithful estimation of the effect size and for which the replication offers additional evidence. This is formulated as θ \approx θ_{orig} with θ_{orig} being the effect size estimate from the original study.

In contrast to the originally proposed Replication Bayes factor, this method estimates the Bayes factor using importance sampling. This yields different results than the code provided by Verhagen and Wagenmakers (2014), but is more stable and less biased than the Monte Carlo estimate (see Bos, 2002) in cases where original and replication study yield different effect size estimates.

Value

An object of ReplicationBF class.

References

• \insertRef

  Verhagen2014ReplicationBF

• \insertRef

  Bos2002ReplicationBF

Examples

## Not run: 
# Example 1 from Verhagen & Wagenmakers (2014)
# Using a Normal approximation to the original's posterior distribution
RBF_ttest(2.18, c(10, 11), 3.06, c(27, 27), method = "NormApprox")
RBF_ttest(2.18, c(10, 11), 0.25, c(27, 27), method = "NormApprox")
RBF_ttest(2.18, c(10, 11), 2.44, c(16, 17), method = "NormApprox")

# Using MCMC to draw samples from the original's posterior distribution
RBF_ttest(2.10, c(11, 11), 3.06, c(27, 28), method = "MCMC")
RBF_ttest(2.18, c(10, 11), 0.25, c(27, 27), method = "MCMC")
RBF_ttest(2.18, c(10, 11), 2.44, c(16, 17), method = "MCMC")

## End(Not run)

neurotroph/ReplicationBF documentation built on May 28, 2019, 3:39 p.m.