Github user actuaryzhang commented on the issue:
https://github.com/apache/spark/pull/16149
@srowen @sethah
Thanks for the comments. Yes, the major use case is to be able to handle
multiple trials (integer weight, real-valued response). Indeed, a better way to
do this is through `offset`, which I have proposed to do in this JIRA
[SPARK-18710](https://issues.apache.org/jira/browse/SPARK-18710). Please let me
know if this is worth pursuing.
I have submitted another two commits.
1. One commit makes minimal modification to the exiting Binomial GLM test
so that one response record is now non-integer. This test still failed because
the deviance residual calculation seems to work only for `y in (0, 1)`
2. The second commit fixes the issue in calculating the deviance. But I
think the code can be improved, especially regarding the function `y_logy`.
What's the best way to create a utility function like this? Please advise.
With the two commits, all tests now passed, including the ones on AIC and
the deviance.
---
If your project is set up for it, you can reply to this email and have your
reply appear on GitHub as well. If your project does not have this feature
enabled and wishes so, or if the feature is enabled but not working, please
contact infrastructure at [email protected] or file a JIRA ticket
with INFRA.
---
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
