News for hackers. Stuff that matters.

A comment on Causal inference as a blind spot of data scientists. Parent comment

Re:Causal inference as a blind spot of data scientists by riedel on Sunday October 15, 2023 @12:49PM (#37891305)
Actually causal inference is also really hard to benchmark. My colleague started an effort to be actually able to reproduce and compare results. Also the algorithms often do not scale too well.

Everytime we wanted to use this for real data it is just a little bit too much effort and the results are not conclusive because it is hard to verify huge graphs. My colleague e.g. wanted to apply it explain risk confounders in investment funds.

I personally also do not like the definition of causality they base it on.

[ Parent | on HN ]
Re:Causal inference as a blind spot of data scientists by kjkjadksj on Sunday October 15, 2023 @01:06PM (#37891470)
You don’t need to look at a graph at all though, right? There are plenty of tests that can help you identify factors that could be significantly affecting your distribution
[ Parent | on HN ]
Re:Causal inference as a blind spot of data scientists by bigfudge on Sunday October 15, 2023 @01:54PM (#37891840)
If you want to make causal inferences you really do have to look at a graph that includes both observed and probable unobserved causes to get any real sense of what’s going on. Automated methods absent real thinking about the data generating process are junk.
[ Parent | on HN ]
Re:Causal inference as a blind spot of data scientists by esafak on Sunday October 15, 2023 @02:50PM (#37892296)
You can only select among features that you have measured.
[ Parent | on HN ]
Re:Causal inference as a blind spot of data scientists by muraiki on Sunday October 15, 2023 @05:31PM (#37893564)
“Graph” here means the directed acyclic graph encoding the causal relationships, not a chart of a distribution.
[ Parent | on HN ]
Re:Causal inference as a blind spot of data scientists by Dzidas on Sunday October 15, 2023 @02:14PM (#37891993)
One way to test this is through a placebo test, where you shift the treatment, such as moving it to an earlier date, which I have seen used successfully in practice. Another approach is to test the sensitivity of each feature, which is often considered more of an art than a science. In practice, I haven't observed much success with this method.
[ Parent | on HN ]
Re:Causal inference as a blind spot of data scientists by esafak on Sunday October 15, 2023 @02:48PM (#37892275)
Go on, please. What definition, and algorithms with scaling problems?
[ Parent | on HN ]