This project is read-only.


A KeepTest is a way to filter out work items either during initial computation or during tabulation.

Useful KeepTests


keep every work item


throw out associations between amino acids of the same position. That is, throws out associations where the pos values are the same.
The amino acid variables in the input files must be of the form [aa]@[pos] or [pos]@[aa], for example, A@301 or 301@A


where p is a double. Keeps p fraction of the associations that would otherwise be kept.
For example, KeepRandom0.001


keeps rows where either pred or targ is var1 or var2, etc.
For example, KeepOnlyA01;C17


complement of KeepOnly


make sure the variables occur at least N times in nodes for which neither variable is missing. The statistics cannot be trusted if N < 2.


all have to be true.


at least one has to be true

KeepFisherOneDirection[Attraction | Escape]

keep only positive (attraction) or negative (escape) associations. Useful for calculating q-values separately in tabulate.


when run from command line, may need double quotes (ie, quotes have to be in the string itself). For example Row[””PredictorVariable””]=""C17""


Only works for PhyloD. This is a special keep test. It requires that the sparse files for pred and targ have the same number of variables. It then reads through both sparse files, and keeps only matching pairs (by position in the file). This allows you to not use the cross product, which saves lots of space/time, with the tradeoff that the sparse files get large. Serves as a flag to use a different WorkList to save computation.

Last edited Mar 22, 2007 at 8:02 PM by CarlK, version 3


No comments yet.