These two approaches do therefore not necessarily overlap in their conclusions as illustrated in this report and as pointed out by others ( Bzdok et al., 2018 ); highly significant result (in the classical sense) provide no guarantees on the (held-out) predictive capabilities of the model so it cannot be excluded that a more marked difference between the two groups (e.g., p -values ≤ 0.001) would have resulted in a higher accuracy.
← all excerpts
Whole-Brain Exploratory Analysis of Functional Task Response Following Erythropoietin Treatment in Mood Disorders: A Supervised Machine Learning Approach.
1
—
—