@replicationindex.com I find that it is more meaningful for a non-expert audience to have it in frequency rather than density so that they can directly compare the obtained versus expected values, but I wanted to know if this is still valid in your opinion or if I am missing something. Thanks!

I will have a look, thanks for the information.

I don't think it shows the same thing. Here we wanted to highlight publication bias by studying the distribution of p-values.
However, the average power of the positive results would definitely have been very useful data for assessing the reliability of the results!

Exactly, whether it's for us practitioners or our patients, a change is needed here.

Special thanks to @pbogdan.bsky.social for making his code available, which allowed us to calculate the proportion of fragile p-values in our sample 🙏

✍️ Conclusion

Questionable research practices, publication bias… Too many grey areas persist around the credibility of physical therapy research, and although tools exist to improve transparency, they still appear to be scarcely used. It’s time to base our practice on (real) evidence.

Preprints:
- Only 6 articles (1%) were posted as preprints, slowing the pace of scientific discovery.

Open research practices:
- No registered reports, which help ensure methodological rigor.
- 72% of studies are not preregistered, leaving room for protocol or outcome changes.
- Only 12% of articles make their data accessible, which is necessary for reproducibility.

Rigorous research practices:
- Sample size was justified in only 37% of quantitative studies, reducing the reliability of the results.
- No replication studies, even though they are essential to confirm the robustness of findings.

Publication bias:
- there are 2.6 times more p-values just below the 0.05 threshold than p-values just above it.
- 39% of p-values in an article fall in the “fragile” zone (.01 < p < .05), whereas a maximum of 26% is expected for studies with 80% statistical power.

📑 Results

High positive result rate:
- more than 8 out of 10 studies report a statistically significant result, an abnormally high and alarming proportion.

⚙️ Methods

465 articles published between 2022 and 2024 in official
journals of national physical therapy associations in the USA, Canada, the UK, and Australia were analyzed.

📰 New preprint: Replicability and transparency in physical therapy research: Time to wake up 📰

Where does physical therapy stand on the replication crisis and open-science practices?

We have identified significant shortcomings in current physical therapy research practice
👉 doi.org/10.1101/2025...

📣 New preprint 📣

The brain attenuates self-touch, but how does this unfold at the neural level before the touch? We used MEG to find out 🧠 👉👈

New Replication Project in sport and exercise science shares it results: Out of 25 replication studies performed, 56% yielded significant results, but only 7 (28%) revealed a similar effect size. Several problems related to the quality of are discussed.

link.springer.com/article/10.1...