Awesome.

Also ANY observational data (i.e. that not obtained from randomised controlled experiments) is likely to be full of known and unknown biases. It would take domain experts to guess what biases are likely to be there, and even they would not think of everything.

On coding.

Domain experts will produce 'terrible' code that just does the required job perfectly and fast too. But will probably be unmaintainable.

Software experts will produce 1000000 lines of object orientated code with multiple dependencies, 90% of which is irrelevant and just about does the job (slowly). This after multiple iterations of missunderstandings about what the software is supposed to do. But I guess the code may be intelligable to other developers (maybe).

Computers have been around for decades now and we haven't solved this problem.