Stahl, C., Klauer, K. C., & Erdfelder, E. (2008). Matching bias in the selection task is not eliminated by explicit negations. Thinking and Reasoning, 14(3), 281–303. doi 10.1080/13546780802116807
When the above article was first published, there were some reporting errors in the article that were due to the use of an incorrect data file version in the original analyses. They are corrected below. The corrections affect neither the pattern of results discussed nor the conclusions drawn in the article. Raw data and additional material can be obtained at http://osf.io/q5ssw.Footnote1
The correct number of participants in the 8 groups of Experiment 2 is 351, 343, 339, 308, 326, 348, 300 and 346 (p. 288). In Table 1 (p. 291), the correct values for AMI, CMI and LI for Experiment 2 are 0.16, 0.20 and 0.37 for the implicit-negation condition, as well as 0.05, 0.10 and 0.31 for the explicit-negation condition (and the correct values of the rescaled indices discussed on p. 295 are, therefore, 0.64 and 0.80). The correct statistics for the -tests against zero for these indices (reported on p. 292) are 7.55, 9.83 and 12.79 (df = 1340, all
) for the implicit-negation condition; and 2.61, 4.41 and 11.09 (df = 1319, all
) for the explicit-negation condition. The correct statistics for the difference between implicit and explicit conditions are, for the AMI,
,
; for the CMI,
,
; and for the LI,
,
(
). The correct effect sizes for AMI and CMI (discussed on p. 295) are
and
, which (assuming
and
) require samples sizes of
and
for detection; given
and
, these effect sizes can be detected with negligible power (0.10 and 0.16). In Table A2 (Appendix), the correct estimates (and 95% CIs) for parameter
in Experiment 1 are (for conditions A3, An3, nA3, nAn3, respectively): 0.79 (0.71, 0.87), 0.73 (0.62, 0.85), 0.55 (0.43, 0.66), 0.74 (0.64, 0.84); and the correct estimates for parameter
in the explicit-negation groups of Experiment 2 are 1 (0, 1), 0.32 (0, 1), 0 (0, 1), 0.45 (0.13, 0.77). None of the above corrections affected the article's substantive conclusions.
Acknowledgment
The authors are grateful to Phil Johnson-Laird for bringing the discrepancies in the reported results to their attention.
Notes
1 The present analyses used R (3.3.1, R Core Team, Citation2016) and the R-packages MPTinR (1.10.3, Singmann & Kellen, Citation2013), papaja (0.1.0.9479, Aust & Barth, Citation2016), snow (Knaus, Citation2015; 0.4.2, Tierney, Rossini, Li, & Sevcikova, Citation2016), and snowfall (1.84.6.1, Knaus, Citation2015).
References
- Aust, F., & Barth, M. (2016). Papaja: Create APA manuscripts with R Markdown. Retrieved from https://github.com/crsh/papaja
- Knaus, J. (2015). Snowfall: Easier cluster computing (based on snow). Retrieved from https://CRAN.R-project.org/package=snowfall
- R Core Team,(2016). R: A language and environment for statistical computing. Vienna: R Foundation for Statistical Computing. Retrieved from https://www.R-project.org/
- Singmann, H., & Kellen, D. (2013). MPTinR: Analysis of multinomial processing tree models in R. Behavior Research Methods, 45(2), 560–575. doi:10.3758/s13428-012-0259-0
- Tierney, L., Rossini, A. J., Li, N., & Sevcikova, H. (2016). Snow: Simple network of workstations. Retrieved from https://CRAN.R-project.org/package=snow