Technology
Running experiments when there are network effects
2020-01-27
Download
Right click and do "save link as"
Traditional A/B tests assume that whether or not one person got a treatment has no effect on the experiment outcome for another person. But that’s not a safe assumption, especially when there are network effects (like in almost any social context, for instance!) SUTVA, or the stable treatment unit value assumption, is a big phrase for this assumption and violations of SUTVA make for some pretty interesting experiment designs. From news feeds in LinkedIn to disentangling herd immunity from individual immunity in vaccine studies, indirect (i.e. network) effects in experiments can be just as big as, or even bigger than, direct (i.e. individual effects). And this is what we talk about this week on the podcast.
Relevant links:
http://hanj.cs.illinois.edu/pdf/www15_hgui.pdf
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC2600548/pdf/nihms-73860.pdf
view more
More Episodes
Rock the ROC Curve
2017-01-30
2
2017-01-30
2
Ensemble Algorithms
2017-01-23
8
2017-01-23
8
How to evaluate a translation: BLEU scores
2017-01-16
1
2017-01-16
1
Zero Shot Translation
2017-01-09
2017-01-09
Google Neural Machine Translation
2017-01-02
2017-01-02
How to Lose at Kaggle
2016-12-12
2
2016-12-12
2
Attacking Discrimination in Machine Learning
2016-12-05
1
2016-12-05
1
Recurrent Neural Nets
2016-11-28
7
2016-11-28
7
Neural Net Cryptography
2016-11-14
2016-11-14
Deep Blue
2016-11-07
2016-11-07
Organizing Google's Datasets
2016-10-31
2016-10-31
Fighting Cancer with Data Science: Followup
2016-10-24
7
2016-10-24
7
The 19-year-old determining the US election
2016-10-17
2
2016-10-17
2
How to Steal a Model
2016-10-09
2
2016-10-09
2
Regularization
2016-10-03
2016-10-03
The Cold Start Problem
2016-09-26
2
2016-09-26
2
Open Source Software for Data Science
2016-09-19
3
2016-09-19
3
Get this podcast on your
phone, FREE
