if we have already calculated the linear predictor using fit$predict, this function can be useful.
"calc.loss.from.lp.list"