- From: Erik Taubeneck via GitHub <sysbot+gh@w3.org>
- Date: Wed, 22 Jun 2022 04:40:51 +0000
- To: public-patcg@w3.org
Thanks for this write up @benjaminsavage. I hope it will help clarify the conversation tomorrow. The only thing I'd add is that along with counts and sums, we often also want variance (of the value.) This is particularly acute in the "conversion-lift" case, where we want to establish a statistically significant difference between the test and control group (but more generally, all these are often used for comparisons, which are more meaningful with a measure of variance.) One small wrinkle here is that we may actually prefer a differential private confidence interval, rather than a differentially private mean (eg sum) and variance, as it may be more efficient (in terms of the privacy budget.) This latter point is probably beyond scope for an MVP, but something else we may want to be able to support in the future. -- GitHub Notification of comment by eriktaubeneck Please view or discuss this issue at https://github.com/patcg/private-measurement/issues/16#issuecomment-1162631660 using your GitHub account -- Sent via github-notify-ml as configured in https://github.com/w3c/github-notify-ml-config
Received on Wednesday, 22 June 2022 04:40:52 UTC