The Use of Balanced Half-Sample Replication in Cross-Validation Studies

摘要: Abstract Cross-validation techniques involve omitting a portion of the available data, fitting prediction function to remaining, and then testing fitted on omitted data. Ideally, one would like repeat this process all possible splits. In many instances, is not computationally feasible. This paper argues demonstrates, using half samples by way illustration, that balanced sampling developed for analysis complex sample survey data provide an efficient population