WIP: Bayes test #2258
WIP: Bayes test #2258
Conversation
openshift-ci
bot
added
the
do-not-merge/work-in-progress
|
[APPROVALNOTIFIER] This PR is APPROVED This pull-request has been approved by: dgoodwin The full list of commands accepted by this bot can be found here. The pull request process is described here
Needs approval from an approver in each of these files:
Approvers can indicate their approval by writing |
openshift-ci
bot
added
the
approved
This is because we don't want heavily run tests to dominate the new samples
The R version came first and essentially uses the dnbinom function. I had AI translate it into Go and it had to do some of the math the hard way. The Go output is nearly identical, just lacking some precision. An idea would be to run trials with various distributions to see which turned out to be the most accurate. What matters in this code is the "Posterior Predictive Probability" output. A low probabily means the prior data doesn't fit the new data well so it might be worth investigating. You can see that setting a threshold of "Flag the test if we're 95% certain something weird is going on" would work in all these examples. In practice, I think we could set it to 99% (Meaning, the Posterior Predictive Probability would be <= 1%). Right now this doesn't feel that different than p-values and Fisher's Exact Number, but I think the nice part is that the math can work with any sample size at all---even a sample of 1. Obviously, the more data that comes in the more confidence we'd have.
Negative binomial PR analysis
|
@dgoodwin: The following tests failed, say
Full PR test history. Your PR dashboard. Instructions for interacting with me using PR comments are available here. If you have questions or suggestions related to my behavior, please file an issue against the kubernetes-sigs/prow repository. I understand the commands that are listed here. |
No description provided.