0.25
0.5
0.75
1.25
1.5
1.75
2
Clustered Graph Randomization: Network Exposure to Multiple Universes
Published on 2013-09-275875 Views
A/B testing is a standard approach for evaluating the effect of online experiments; the goal is to estimate the `average treatment effect' of a new feature or condition by exposing a sample of the ove
Presentation
Clustered Graph Randomization: Network Exposure to Multiple Universes00:00
A/B testing a social product change - 102:37:44
A/B testing a social product change - 211:23:23
A/B testing a social product change - 321:56:01
A/B testing a social product change - 433:23:20
A/B testing a social product change - 536:56:19
A/B testing a social product change - 640:31:31
Measuring ‘Average Treatment Effect’56:09:36
Average Treatment Effect: what changes?95:22:17
Agenda124:18:45
Defining ‘network exposure’ - 1155:52:14
Defining ‘network exposure’ - 2156:33:11
Defining ‘network exposure’ - 3186:12:54
Clustered graph randomization217:47:27
How to partition the graph? - 1220:44:32
How to partition the graph? - 2228:26:39
Cluster and randomize235:17:12
Cluster and randomize... finely.237:12:01
How to cluster? What algorithm?245:50:26
Probabilities under clustered coin flips - 1281:02:51
Probabilities under clustered coin flips - 2293:54:25
Characterize ATE Variance299:37:30
ATE Variance301:48:28
Cycle graph example - 1309:25:02
Cycle graph example - 2312:46:23
Cycle graph example - 3321:18:39
Cycle graph example - 4322:11:11
Cycle graph example - 5327:05:25
Cycle graph example - 6329:08:48
Cycle graph example - 7329:49:31
Cycle graph example - 8335:39:37
Restricted growth condition342:53:51
Restricted growth graphs - 1343:07:50
Restricted growth graphs - 2348:41:33
Restricted growth graphs - 3360:48:24
Clustering restricted growth graphs369:48:06
Conclusions - 1380:17:55
Conclusions - 2380:34:14