<But how do you even _define_ the 
loss function for a discrete outcome? Hmmm... mumble, mumble...

Something on binary outcomes should be here

http://www.albany.edu/economics/research/workingp/2012/lahiriyang.pdf

Oleh