The paper extends seven methods for constructing confidence intervals that are developed for the estimation of binomial proportion with continuity correction to the estimation of multinomial proportion. The Bonferroni method is used to limit the overall confidence coefficient to ? %. The study considers each cell versus the remaining cells as a binomial estimate for the individual cell proportion (Cochran's, 1963). An analytical study is performed, for = 5, 10, 15, 20, 30 and 40; for number of categories limited to five, and for 5% and 1%. Results for the seven extended methods were checked for the invariance property and interval width. Results show that all methods satisfy invariance only for large n; the Blyth and Still method and the Wilson Score Method (Fleiss (1981) prove to do well in the estimation of multinomial proportion for all sample sizes; these two methods satisfy invariance property; however, the interval widths produced are not the narrowest. Hall Method gives the narrowest intervals, but it does not satisfy invariance. The Clopper-Pearson Exact Method satisfies invariance under some conditions when applied to > 1, and the interval widths area bit higher than Method of Hall.