Exercise 6.2#
(a)#
(i) False. Lasso reduces the number of variables, thus is less flexible.
(ii) False. Same justification as in (i).
(iii) True.
(iv) False. In general, lasso reduces variance and increases bias. Reduction in variance should compensate increasement in bias.
(b)#
Same answer as above.
(c)#
(i) False. In general, non-linear methods reduce bias and increase variance. Reduction in bias should be compensate increasement in variance.
(ii) True.
(iii) False. Non-linear methods are more flexible because they accomodate to the data. Unlike least squares, non-linear methods don’t assume a parametrized relationship between the predictors and the response.
(iv) False. Same justification as in (iii).