Confidence intervals for the variance and the ratio of two variances of non-normal distributions with missing data

Abstract

This paper compares confidence intervals for the variance and the ratio of two variances when the population distributions are non-normal and item nonresponse is occurring. The data after random hot deck imputation used to define the confidence interval. The confidence intervals considered are the classical confidence intervals in text books and the adaptive confidence interval based on the Bonett confidence intervals. Our simulation study shows that the use of the adaptive confidence intervals for variance and ratio of two variances when the underlying distributions are generally skewed and unknown and missing data occur give better coverage probabilities. Therefore their use is recommended.