We introduce a novel method of sensitivity analysis to aid in assessing the missing-at-random assumption in a two-way contingency table with one supplemental margin. The procedure involves a set of response odds and nonresponse odds computed from the observed counts.