A purist would argue that analytical integrity requires refusal; issue an ultimatum and resign if necessary. Can less than complete refusal ever be ethical? We should keep in mind that, because analysis involves prediction, analysts rarely enjoy complete confidence in their conclusions. Careful analysts check the sensitivity of their results to changes in critical assumptions and convey the level of confidence they have in their conclusions to their clients. We can imagine analysts developing plausible ranges of results. For example, although the analyst believes that the cost of some program is likely to be close to $10 million, the most conservative assumptions might lead to an estimate of $15 million and the most optimistic assumptions to an estimate of $5 million. After making the range and best estimate clear to the client, would it be ethical for the analyst to prepare a version of the analysis for public distribution that used only the most optimistic assumptions?