search

anujkhare / iregnet

7 stars 12 forks source link

Deviance #17

Open anujkhare opened 8 years ago

anujkhare commented 8 years ago

Deviance is just the residual taken to the log-likelihood scale.

tdhock commented 8 years ago

I think the deviance is just the negative log-likelihood (aka the loss function), which you can use as a model selection criterion for the cv.iregnet function (minimize the deviance on a held-out validation set)