When I was deep in the depths of comparing methods for computer certification of verbal autopsies, I was a big fan of Random Forests for determining underlying cause of death from interview data. But my colleagues urged me to consider methods that would be easier to inspect and to explain. I’m starting to see why, for example the exchange below:
Sent: Tuesday, March 15, 2016 7:48 AM
To: Abraham D. Flaxman
I hope everything is going well with you.
We are improve the translation of the short VA to Portuguese, specifically, the questions
adult_2 7 and
adult_2_10. Would you send us the tariffs of these symptoms? They will help us so much.
2016-03-15 17:50 GMT-03:00 Abraham D. Flaxman :
Happy to help:
Can I publish your question and my answer on my blog https://healthyalgorithms.wordpress.com ? I think this sort of thinking will be helpful for others doing VA translation work.