Dear All,
Is there a theory allowing harmless drop of empty cells in a contingency table? E.g. table(v1,v2,v3,v4) applied to iris flower data produces a large 4D table with considerable number of empty cells. My wish is to eliminate levels for which crossings are empty. Dropping rows or columns or layers or strata which sum to 1 would be acceptable? Hopefully I do not hurt you, statisticians.
Thanks in advance,
Ray Haraf.
|