When comparing models containing a different number of predictors in the estimation, we may obtain an estimate of test error using C subscript p, A I C, B I C, or A d j u s t e d space R squared which apply a penalty based on the number of used predictors d, to a. training error using mu subscript y. b. test error using M S E. c. sample error using fraction numerator n minus 1 over denominator n end fraction. d. training error using R S S