F. Kianifard et Wh. Swallow, A REVIEW OF THE DEVELOPMENT AND APPLICATION OF RECURSIVE RESIDUALS INLINEAR-MODELS, Journal of the American Statistical Association, 91(433), 1996, pp. 391-400
Recursive residuals have been shown to be useful in a variety of appli
cations in linear models. Unlike the more familiar ordinary least squa
res residuals or studentized residuals, recursive residuals are indepe
ndent as well as homoscedastic under the model. Their independence is
particularly appealing for use in developing test statistics. They are
not uniquely defined; their values depend on the order in which they
are calculated, although their properties do not. In some applications
one can exploit this order dependence, coupled with the fact that the
y are in clear one-to-one correspondence with the observations for whi
ch they are calculated. Uses for recursive residuals have been suggest
ed in almost all areas of regression model validation. Regression diag
nostics have been constructed from recursive residuals for detecting s
erial correlation, heteroscedasticity, functional misspecification, an
d structural change. Other statistics based on recursive residuals hav
e focused on detection of outliers or observations that are influentia
l or have high leverage. Recent work has explored properties and possi
ble uses of recursive residuals in models with a general covariance ma
trix, multivariate linear models, and nonlinear models. Computing rout
ines are available for obtaining recursive residuals accurately and ef
ficiently.