View Single Post
Old 04-03-2013, 04:10 PM
Michael Reach Michael Reach is offline
Senior Member
Join Date: Apr 2013
Location: Baltimore, Maryland, USA
Posts: 71
Default Family-wise error rate

I am trying to translate some of what I heard in Lecture 2 to things that sounded similar from statistics. Slide 16/17 ("A simple solution"), where we just add up all the probabilities of error, reminds me of what they call the Bonferroni correction. I guess it's considered safe, but much too conservative. Generally they use a couple of other techniques, like controlling the False Discovery Rate. You don't try to keep the sum of all the errors down, but settle for bringing most of them down on average.
I think the professor actually hinted in the lecture that the error estimate can be improved. I'm really just posting this as an attempt to assimilate what I know from elsewhere and what I'm learning here.
Reply With Quote