Problem set 8
Complete this task in teams of two or three students.
Submission information: please submit on ZoneCours
- a PDF report
- your code
Look at the ANCOVA example; this should be helpful in completing the problem set.
We consider Experiment 3 of Stekelenburg et al. (2021); (download the paper). The database can be found in SSVB21_S3
in the hecedsm
package. You can also download the SPSS database via this link.
- Have a quick look at the exclusion guidelines in the preregistration. Do they make sense? think about other potential criteria that could have been listed.
- Write down potential improvement points related to statistics and experimental setup that you could raise if you were assigned to peer-review the paper.
- The author proceed with splitting the data set in three groups and performing each pairwise comparisons in turn, corresponding to the different hypotheses (circa lines 499-533 of the script). Why might this approach be suboptimal relative to the one that consists in fitting a model to the three categories (pooled variance) and proceeding later with computing pairwise differences and contrasts?
- Test the assumptions of
- equal variance per experimental condition with the ANCOVA
- proper randomization based only on
Prior
scores - equality of slopes
- If the variance are unequal, fit the model with unequal variances per group (see the ANCOVA example) — otherwise proceed as usual.
- Look at the pairwise difference between between Boost+ and consensus only condition and report the results of this test.
- Does the conclusion of the test change relative to that reported in the paper?
References
Stekelenburg, A. van, Schaap, G., Veling, H., & Buijzen, M. (2021). Boosting understanding and identification of scientific consensus can help to correct false beliefs. Psychological Science, 32(10), 1549–1565. https://doi.org/10.1177/09567976211007788