Direct Estimation of Confidence Intervals for Proportion by Means of Continuity Simulation of Binomial Distribution

Niuman M Comas

Departamento de Matemática, Universidad de Holguín, Ave XX Aniversario, 82100, Holguín, Cuba.

Abstract:

This paperexplains a methodology to obtain confidence intervals for proportion without any normal approximation. It is based on the use ofbinomial distribution in order to calculate the confidence bounds using polynomial interpolation techniques to estimate “artificial” probabilities associated to fractional distribution arguments.

The methodology includes the estimation of cumulative probabilities with continuity simulation and the respective inverse root. Hypothesis test is also incorporated. Abrief review of other common methods is previously shown.

The implementation of the “Direct” method wascarried outunder open source by means of JavaScript programming and it is available in the on-line statistical laboratory DynStats.com.

The results were compared with other procedures like the Wald, Clopper-Pearson and Mid-P methods. Most similarities were found for the Wald and Mid-p methods according to the proportion’s region.

Keywords: Confidence interval, binomial proportion, binomial distribution, polynomial interpolation.

Pages: 110 – 122 | Full PDF Paper