BOOTSTRAP-BASED INFERENCE FOR GROUPED DATA

Abstract

Grouped data refers to continuous variables that are partitioned in intervals, not necessarily of the same length, to facilitate its interpretation. Unlike in ungrouped data, estimating simple summary statistics as the mean and mode, or more complex ones as a percentile or the coefficient of variation, is a difficult endeavour in grouped data. When the probability distribution generating the data is unknown, inference in ungrouped data is carried out using parametric or nonparametric resampling methods. However, there are no equivalent methods in the case of grouped data. Here, a bootstrap-based procedure to estimate the parameters of an unknown distribution based on grouped data is proposed, described and illustrated.