replicationInterval-package: Replication Interval Functions

In replicationInterval: Replication Interval Functions

Description

A common problem faced by journal reviewers and authors is the question of whether the results of a replication study are consistent with the original published study. One solution to this problem is to examine the effect size from the original study and generate the range of effect sizes that could reasonably be obtained (due to random sampling) in a replication attempt (i.e., calculate a replication interval). If a replication effect size falls outside the replication interval, then that effect likely did not occur due to the effects of sampling error alone. Alternatively, if a replication effect size falls within the replication interval, then the replication effect could have reasonably occurred due to the effects of sampling error alone. This package has functions that calculate the replication interval for the correlation (i.e., r), standardized mean difference (i.e., d-value), and mean. The calculations used in version 2.0.0 and onward differ from past calculations due to feedback during the journal review process. The new calculations allow for a more precise interpretation of the replication interval.

Package:	replicationInterval
Type:	Package
Version:	2.0.1
Date:	2016-05-24
License:	MIT License + file LICENSE

Details

ri.r creates a replication interval for a correlation (i.e., r )
ri.d creates a replication interval for a standardized mean difference (i.e., d )
ri.m creates a replication interval for a mean (i.e., M )

ri.r.demo demonstrates RI capture percentage for a correlation (i.e., r )
ri.d.demo demonstrates RI capture percentage for a standardized mean difference (i.e., d )
ri.m.demo demonstrates RI capture percentage for a mean (i.e., M )

Author(s)

Author:	David J. Stanley [e-mail]
Maintainer:	David J. Stanley [e-mail]

References

Spence, J.R. & Stanley, D.J.(in prep). Replication Interval: What to expect when you're expecting a replication.


Also:

Cumming, G. & Maillardet, R. (2006). Confidence intervals and replication: where will the next mean fall? Psychological Methods, 11(3), 217-227.

Estes, W.K. (1997). On the communication of information by displays of standard error and confidence intervals. Psychonomic Bulleting & Review, 4(3), 330-341.

Zou, G.Y. (2007). Toward using a confidence intervals to compare correlations. Psychological Methods, 12(4), 399-413.

Examples

ri.r(r=.35,n=100,rep.n=200)
ri.d(d=.65,n1=50,n2=50,rep.n1=100,rep.n2=100)
ri.m(M=2.53,SD=1.02,n=40,rep.n=80)

Example output

Original study: r = 0.35, N = 100, 95% CI[0.16, 0.51]
Replication study: N = 200
Replication interval: 95% RI[0.13,0.56].


Interpretation:
The original correlation is 0.35 - with a replication interval 95% RI[0.13, 0.56] based a replication sample size of N = 200. If the replication correlation differs from the original correlation only due to sampling error, there is a 95% chance the replication result will fall in this interval. If the replication correlation falls outside of this range, factors beyond sampling error are likely also responsible for the difference.
Original study: d = 0.65, N1 = 50, N2 = 50, 95% CI[0.25, 1.05]
Replication study: N1 = 100, N2 = 100
Replication interval: 95% RI[0.16,1.14].


Interpretation:
The original d-value is 0.65 - with a replication interval 95% RI[0.16, 1.14] based a replication cell sizes N1 = 100 and N2 = 100. If the replication d-value differs from the original d-value only due to sampling error, there is a 95% chance the replication result will fall in this interval. If the replication d-value falls outside of this range, factors beyond sampling error are likely also responsible for the difference.
Original study: M = 2.53, SD = 1.02, N = 40, 95% CI[2.20, 2.86]
Replication study: N = 80
Replication interval: 95% RI[2.13,2.93].


Interpretation:
The original mean is 2.53 - with a replication interval 95% RI[2.13, 2.93] based a replication sample size of N = 80. If the replication mean differs from the original mean only due to sampling error, there is a 95% chance the replication result will fall in this interval. If the replication mean falls outside of this range, factors beyond sampling error are likely also responsible for the difference.

replicationInterval documentation built on May 2, 2019, 3:35 p.m.