Re:BayesDB: Data science is a communication problem
by mcbrown on Thursday March 31, 2016 @08:47PM
(#11401776)
BayesDB and BQL are great ideas in theory. But the problem with building any "general" analysis tool to abstract away underlying complexity is that it must either be (a) a least common denominator solution, or (b) the proverbial hammer to which everything will look like a nail.
The reason why there are lots of different statistical methods is that different problems and different samples from different populations call for different approaches. And frequently the reason why a certain approach is invalid in a particular use case is subtle, difficult to explain, and easy to miss. Abstracting that complexity away from the would-be "scientist" is an invitation for them to develop unreasonable confidence in their results "because the model said so." We have more than enough of that attitude already, thank you very much.
Re:BayesDB: Data science is a communication problem
by tryitnow on Friday April 01, 2016 @11:14AM
(#11405372)
I think there's potential here, but the claims made in this article don't really hold for me.
I have no idea why BQL would help facilitate communication between domain experts, business analysts, and data scientists.
The parallel to SQL is telling:
Frankly, almost no domain experts waste their time on SQL and most business analysts have better tools suited for their common use cases. So why would these two groups get much out of BQL?
Plus it's another thing for data scientists to learn?
The three stakeholders mentioned in this article don't seem to really benefit from it.
Why not just use R? It solves the problem of communication with non-technical stakeholders through tools like Markdown, knitr, shiny, etc. It has a relatively long history of use, a supportive community, it's super easy to look up how to do things, etc.
I'm sure there are good reasons for BQL, but this article doesn't really summarize them.
The reason why there are lots of different statistical methods is that different problems and different samples from different populations call for different approaches. And frequently the reason why a certain approach is invalid in a particular use case is subtle, difficult to explain, and easy to miss. Abstracting that complexity away from the would-be "scientist" is an invitation for them to develop unreasonable confidence in their results "because the model said so." We have more than enough of that attitude already, thank you very much.
I have no idea why BQL would help facilitate communication between domain experts, business analysts, and data scientists.
The parallel to SQL is telling:
Frankly, almost no domain experts waste their time on SQL and most business analysts have better tools suited for their common use cases. So why would these two groups get much out of BQL?
Plus it's another thing for data scientists to learn?
The three stakeholders mentioned in this article don't seem to really benefit from it.
Why not just use R? It solves the problem of communication with non-technical stakeholders through tools like Markdown, knitr, shiny, etc. It has a relatively long history of use, a supportive community, it's super easy to look up how to do things, etc.
I'm sure there are good reasons for BQL, but this article doesn't really summarize them.
Still alive, as far as I know.