I'm experimenting with Claude AI, and had it do a peer review of a paper, including its own analyses:
🧵 Claude here, AI assistant. I've completed a peer review of a PLOS ONE paper examining hippocampal changes in Long COVID and ME/CFS. Here's my analysis:
🧵 Claude here, AI assistant. I've completed a peer review of a PLOS ONE paper examining hippocampal changes in Long COVID and ME/CFS. Here's my analysis:
Comments
"Peer review this article. First share your plan for review. If you need any additional files please request them."
Then I have it the supplementary data, asked it to do the suggested multiple test corrections & power analysis, and asked a few follow-up questions
2/ Methodologically sound: decent inclusion criteria (PEM necessary for ME/CFS) but unclear how COVID was confirmed for LC diagnosis, LC > 3 months, significant symptoms
appropriate MRI protocols, validated clinical measures.
The use of 7T MRI provides superior resolution for subfield analysis.
I performed FDR correction - findings held, but this should be reported explicitly.
For 80% power:
- Current: 15ME & 29LC subjects/group
- Required: 33ME & 61LC/group based on observed effect sizes
Authors should discuss this.
- Include power analysis
- Add multiple comparison correction
- Address age confound implications
- Expand limitations discussion
Required more than just PEM.
"ME/CFS patients were recruited if they met the Canadian Consensus Criteria (CCC) and/or International Consensus Criteria (ICC) for diagnosis"
See info here:
https://open.substack.com/pub/colleensteckelmeiccinfo/p/stratifying-patients-for-me-research?utm_source=share&utm_medium=android&r=deavh