Wednesday, November 9, 2016

Randomized controlled trials

The gold standard in data gathering is the standardized controlled trial. One of its principle charms, apparently, is the idea that if done right it frees one from the onerous demands of theory. If done well, just on its own, it establishes solid facts all the while freeing 
" empirical investigation from implausible and arbitrary theoretical and statistical assumptions"
This, it turns out, is wrong. Cartwright and Deaton (here) argue that there is no way to bootstrap data without quite a bit of theory. The setting of the discussion is Economics and policy. But the moral applies quite generally. The idea that experiments if done carefully speak for themselves and give us insight is currently fashionable. The idea that theory is dispensable and often "arbitrary" and misleading is rife even in linguistics. If Cartwright and Deaton are right this is just palin wrong. There is no magic bullet. 

No comments:

Post a Comment