SciPost Submission Page
Multiverse Analysis of the Effect of a Non-Deceptive Placebo on a Neural Measure of Emotional Distress
by Hannah E. Fowles, Peter J. Allen
Submission summary
| Authors (as registered SciPost users): | Peter Allen |
| Submission information | |
|---|---|
| Preprint Link: | scipost_202506_00035v2 (pdf) |
| Date submitted: | Nov. 4, 2025, 7:42 p.m. |
| Submitted by: | Peter Allen |
| Submitted to: | Journal of Robustness Reports |
| Ontological classification | |
|---|---|
| Academic field: | Multidisciplinary |
| Specialties: |
|
| Approach: | Experimental |
Abstract
An objective, neurological effect of a non-deceptive placebo observed in [1] is robust to age sub-sampling and electrode selection. However, it seems likely to be a false positive.
Author comments upon resubmission
Thank you to both reviewers for their comments on the first iteration of this manuscript. We have substantially revised the manuscript in response to these. A summary list of changes is provided below. Regards, Peter Allen and Hannah Fowles
List of changes
- We have made the target effect more explicit in the introduction, by including all relevant statistics.
- We have included two references for multiverse, including a very recent paper by Heyman et al. (2025).
- We have removed the N = 100 analytic alternatives. A full rationale for this is provided in our response to R2.
- We have dropped references to practical/clinical significance throughout and re-framed the conclusions to emphasise how (a) the effect sizes are robust to the age and hemisphere alternatives; (b) sticking to the original sampling plan combined with modest Type I error control changes the conclusion one draws about the reliability of the effect; (c) the effect is an outlier relative to others in this space; and (d) due to (b) and (c), it seems likely to be a false positive. Of course, a high-quality replication is needed before a firm conclusion can be drawn about (d), which is the point we now finish on.
- We’ve substantially simplified the graph. Specifically, we’ve stripped it back to unstandardized effect sizes and their confidence intervals, along with flags to indicate whether they cross three reasonable/justifiable significance thresholds.
- We have also made a number of more modest changes, as detailed in our responses to each reviewer.
Current status:
Voting in preparation