Abstract
Pitfalls of approximate methods in computing confidence limits for potency measurements in biological assays are discussed. Determinations of exact confidence limits are described for the log-dose for a given graded response, for a given all-or-none effect, for potency ratios and for assays designed factorially.