Moritz GersterWhy is it so difficult to reproduce published findings?
Is it differing data processing pipelines or differences in the studied cohorts?
Using 1 processing pipeline, I analyzed 5 datasets on #deepbrainstimulation in Parkinson’s disease.
YouTube
Pro: I reproduced key findings by pooling all cohorts.
Con: Reproduction failed when testing hypotheses on individual datasets.
Takeaway: We need larger and more diverse cohorts to maximize the generalizability of our findings.
Check my talk from #NeuralTraces2024 for a summary of my multi-dataset comparison and an exploration of the significance of theta and low gamma oscillations in #ParkinsonsDisease.