
    o  AIC(..., "classical") as default???

    o  use extractAIC instead of AIC

    o  adjust control$M when subsetting
    
    o  allow dfbase vector (including linear terms with df = 1)

    o  names in predict and fitted

    o  multivariate responses

    o  reweight generic?

    o  check loss fuer Poisson!

    o  hatvalues + AIC geht so nur fuer GaussReg!
    o  hat matrices fuer GAMS (Hastie & Tib)
    o  check hatvalues, hatmatrix for boosting with non-gaussian family

    o  weighted median for offset in Huber / Laplace, see Hmisc:::wtd.quantile

