In a subset selection in a regression context, forward selection may not always result in the best subset selection for a given number of parameters in the model. From the answers below choose the one that does not apply to forward selection. Once chosen, a variable that is in a k-variable model can only be replaced if it becomes highly significant when the number of parameters increases form k to k+1. Forward selection chooses the best new single variable at each augmentation where `best' refers to the considered metric (R-square, MSE, BIC, etc.). Forward selection can be used for data with more predictors than observations. Once chosen, a variable remains in the model when the number of parameters increases form k to k+1