🚀 go-pugleaf

Backward elimination does not at all dominate full model fitting.
See:

@article{hou90pre,
   author = {{Van Houwelingen}, J.C. and {Le Cessie}, S.},
   journal = "Statistics in Medicine",
   pages = {1303-1325},
   title = {Predictive value of statistical models},
   volume = {8},
   year = {1990},
   annote = {predictive accuracy; validation; optimism; maximum likelihood}
}

@article{cop83reg,
   author = {Copas, J.B.},
   journal = "Journal of the Royal Statistical Society B",
   pages = {311-354},
   title = {Regression, prediction and shrinkage (with discussion)},
   volume = {45},
   year = {1983},
   annote = {shrinkage; validation; logistic model}
}

@article{cop87cro,
   author = {Copas, J.B.},
   journal = "Journal of the Royal Statistical Society B",
   pages = {175-183},
   title = {Cross--validation shrinkage of regression predictors},
   volume = {49},
   year = {1987},
   annote = {shrinkage; validation; calibration}
}


-- 
----------------------------------------------------------------------------
Frank E Harrell Jr			feh@biostat.mc.duke.edu
Associate Professor of Biostatistics
Division of Biometry                    Duke University Medical Center