week5 additional notes#

Principal component regression is especially used to overcome the problem with collinearity in linear regression by combining explanatory variables to a smaller set of uncorrelated variables.

pcr pcr pcr

Main idea of PCR#

The linear regression model:

\[ Z_i = m(X_i) + \epsilon_i = \beta^TX_i + \epsilon_i \]

degenerates to:

\[ Z_i = \beta^T \Gamma Y + \epsilon_i = m_PC (Y)\Gamma + \epsilon_i \]

Boston housing data Example#