search

mbnebel / DeconfoundedFMRI

DRTMLE to address sampling biases in neuroimaging
MIT License
0 stars 1 forks source link

Compare DRTMLE results run on the cluster to local results #4

Open mbnebel opened 2 years ago

mbnebel commented 2 years ago
  1. check warnings
  2. min propensities estimated by superlearner across all seeds, min propensity ranged from 0.23-0.52
  3. consistency of results across first and second set of 200 seeds: a. correlation of zstats across all edges b. at FDR=.20, name and number of edges indicated to show a difference between diagnosis groups c. at FDR=.05, name and number of edges indicated to show a difference between diagnosis groups
  4. results averaged across all 400 seeds: a. ~at FDR=.20, name and number of edges indicated to show a difference between diagnosis groups~ b. ~at FDR=.05, name and number of edges indicated to show a difference between diagnosis groups~ c. ~number of edges for which DRTMLE mean difference increased compared to naive approach~ d. ~number of edges for which DRTMLE standard errors decreased compared to naive approach~