流行病學研究所;Graduate Institute of EpidemiologyTAI, JOHN, JENJOHN, JENTAI2008-09-262018-06-292008-09-262018-06-292003http://ntur.lib.ntu.edu.tw//handle/246246/81997A family of simultaneous confidence intervals (SCIs) for multinomial proportions is proposed by inverting the power- divergence statistics and the best SCIs in the family is determined, by Monte-Carlo technique. Numerical comparisons of this method with the other alternatives are presented. Simulation results indicate that the new procedure is preferable to all its competitors in most cases.en-UScoverage probabilityMonte-Carlo methodpower-divergence statisticsimultaneous confidence intervalssparse datajournal article