0.25
0.5
0.75
1.25
1.5
1.75
2
Selective Inference and the False Discovery Rate
Published on Jan 31, 20171192 Views
Related categories
Chapter list
Selective Inference and the False Discovery Data00:00
Outline - 100:46
Prolog01:05
The first data-mining problem in statistics - 102:12
The first data-mining problem in statistics - 203:00
The first data-mining problem in statistics - 304:10
The lethal combination of the two05:24
Not only jelly beans06:26
Not only colors06:48
Selective inference08:05
Some notations before we continue08:58
FWER protection09:54
Same for confidence intervals10:57
Old and trusted solutions - 111:29
Old and trusted solutions - 212:38
Old and trusted solutions - 313:14
Significance of 8 strain differences15:28
Unadjusted vs simultaneous16:35
The increasing scale - 117:09
The increasing scale - 217:36
A common feature of the larger applications18:34
The increasing scale changes the goal19:06
Outline - 220:10
The false discovery rate (FDR) criterion20:17
Does it make sense?21:05
Reflections on goals - 122:04
Reflections on goals - 222:32
FDR controlling procedures23:18
Significance of 8 strain23:47
The graphical way to look at it25:09
FDR controlling procedures - adjusted p-values25:42
FDR control of the BH procedure26:25
Positive dependency27:04
Adaptive procedures that control FDR27:38
The graphical approach of Schweder and Spjotvoll28:40
Option 3: The step-down multi-stage procedure28:42
The step-down multi-stage procedure29:28
Bayesian and empirical Bayes approaches30:21
Weighted FDR31:45
FDR a thing of the past?32:46
Selective inference, the false discovery rate and analysis of neuro data33:22
Outline33:33
One concern - different directions33:39
20 parameters to be estimated with 90% CIs34:56
The false coverage-statement rate (FCR)36:53
FCR adjusted selective CIs37:33
Massive selection - by a table39:02
Main table39:22
Odds ratio point and CI estimates - 140:18
Odds ratio point and CI estimates - 240:39
How well do we do?40:59
Adjusting to the selection procedure used42:08
Graphs43:02
Conditional quasi-conventional CI43:55
Example - 144:22
Example - 245:39
Addressing voodoo correlations47:25
Addressing in-study voodoo correlations50:01
Maximum conditional likelihood estimator50:50
Outline - 351:41
Motivation: Wavelets51:49
Testimation53:52
Testimation - same theory55:02
What have we further learned from theory56:38
Wider implications for model selection57:11
What is gained by introducing FDR penalty59:25
Multiple stage FDR01:00:02
Example 2: High dimensionality01:01:14
Rupin´s lab method01:02:13
Separate vs joint FDR testing of families01:02:43
Efron´s comment01:03:58
Justifications for separate FDR testing01:04:49
Selection adjusted separate testing of families01:05:32
Selection adjusted separate testing01:06:13
There was no restriction on the selection rule01:06:56
Hierarchical BH testing01:07:59
Re-analysis of SNP-voxel data for Alzhiemer - 101:08:23
Re-analysis of SNP-voxel data for Alzhiemer - 201:08:27
Other examples - 101:09:22
Other examples - 201:10:55
Multiple phenotypes01:12:33
Selective inference challenges in open access data01:15:28
A potential outcome of this successful summer school01:16:15