Premium
An example of ridge regression difficulties
Author(s) -
Smith Gary
Publication year - 1980
Publication title -
canadian journal of statistics
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.804
H-Index - 51
eISSN - 1708-945X
pISSN - 0319-5724
DOI - 10.2307/3315233
Subject(s) - multicollinearity , normalization (sociology) , bayesian probability , ambiguity , regression , ridge , simple linear regression , mathematics , econometrics , regression analysis , computer science , statistics , geography , sociology , anthropology , programming language , cartography
A simple consumption function is used to illustrate two fundamental difficulties with ridge regression and similarly motivated procedures. The first is the ambiguity of multicollinearity measures for judging the data's “ill‐conditioning”. The second is the sensitivity of the estimates to the arbitrary normalization of the model. Neither of these poses a problem for least squares or Bayesian estimates. The logical restructuring of ridge procedures to avoid these difficulties leads to a more explicitly Bayesian approach.