PSTAT 131 STUDY GUIDE ANSWERS AND
QUESTIONS SET A+
✔✔Regression Spine is LESS flexible than - ✔✔Piecewise polynomial
✔✔# of trees is a tuning parameter for - ✔✔both bagging and boosting
✔✔Bagging - ✔✔make multiple trees and combine (bootstrap)
✔✔Boosting - ✔✔results from one tree feed into another, one by one
✔✔Random Forest classifier used bagged trees that split on random subsets of m<p
predictors to - ✔✔uncorrelate bagged trees.
✔✔In k means clustering, k= - ✔✔# of clusters
✔✔Hierarchical clustering is - ✔✔a bottom up aproach, no need to specify # of clusters,
and initialized by taking each observation as its own cluster.
✔✔Principle component analysis loadings are - ✔✔orthogonal
✔✔Var explained by PC1 - ✔✔is greater than or equal to var explained by PC2
✔✔PVE is useful for - ✔✔choosing # of PCs
✔✔SVM - ✔✔supervised method, has tuning parameters, can be fit using only the inner
product between observations.
✔✔as lambda increases: - ✔✔Hi bias, low var
✔✔as lamdba decreases - ✔✔low bias, hi var
✔✔for a k fold cross validation... - ✔✔larger k = smaller bias, smaller k = larger bias
QUESTIONS SET A+
✔✔Regression Spine is LESS flexible than - ✔✔Piecewise polynomial
✔✔# of trees is a tuning parameter for - ✔✔both bagging and boosting
✔✔Bagging - ✔✔make multiple trees and combine (bootstrap)
✔✔Boosting - ✔✔results from one tree feed into another, one by one
✔✔Random Forest classifier used bagged trees that split on random subsets of m<p
predictors to - ✔✔uncorrelate bagged trees.
✔✔In k means clustering, k= - ✔✔# of clusters
✔✔Hierarchical clustering is - ✔✔a bottom up aproach, no need to specify # of clusters,
and initialized by taking each observation as its own cluster.
✔✔Principle component analysis loadings are - ✔✔orthogonal
✔✔Var explained by PC1 - ✔✔is greater than or equal to var explained by PC2
✔✔PVE is useful for - ✔✔choosing # of PCs
✔✔SVM - ✔✔supervised method, has tuning parameters, can be fit using only the inner
product between observations.
✔✔as lambda increases: - ✔✔Hi bias, low var
✔✔as lamdba decreases - ✔✔low bias, hi var
✔✔for a k fold cross validation... - ✔✔larger k = smaller bias, smaller k = larger bias
