Premium
Regression Shrinkage Methods and Autoregressive Time Series Prediction
Author(s) -
Copas J. B.,
Jones M. C.
Publication year - 1987
Publication title -
australian journal of statistics
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.434
H-Index - 41
eISSN - 1467-842X
pISSN - 0004-9581
DOI - 10.1111/j.1467-842x.1987.tb00744.x
Subject(s) - autoregressive model , shrinkage , regression , statistics , regression analysis , series (stratigraphy) , econometrics , mathematics , time series , partial least squares regression , mean squared prediction error , star model , setar , linear regression , autoregressive integrated moving average , paleontology , biology
Summary The pros and cons of applying regression shrinkage prediction arguments and methods to autoregressive time series forecasting are discussed. Simulation evidence of the performance of a Stein regression prediction formula suggests that the overall dominance of the shrunken predictor over least squares in regression no longer holds in time series samples of a reasonable length. Rather, shrinkage appears the better of the two, with respect to prediction mean squared error, only for weaker relationships and seems to be inferior to the least squares predictor when the autoregressive relationship is strong.