Memory Use. Testing at AIPL indicates that BESTPRED uses about 128 MB of RAM when running. Memory consumption should not grow as a function of the number of records processed unless there is an undetected memory leak. If such behavior does occur please file a bug report.
Batch Processing. Batch processing efficiency can be improved by blocking records by breed. When there is a change in breed within a file the correlations among test days, as well as means and SD of yield, must be recalculated. Sorting the input file by breed will avoid unnecessary calculations resulting from back-and-forth changes in breed codes.
Editing Records. The AIPL in-house processing system includes a very complex series of edits that are performed before records are passed to BESTPRED. These edits catch, for example, overlapping LER segments in input records. Those edits are not replicated in the best prediction programs, although BESTPRED will identify some errors and print warnings to STDOUT (if output is not suppressed). It is essential that users check their logfiles to catch problems arising from problems with input records. BESTPRED may process a record and still return an incorrect answer if there are problems with the input data.See About this document... for information on suggesting changes.