Finite Mixtures Model Approach to Sensitive Questions in Surveys
Finite Mixtures Model Approach to Sensitive Questions in Surveys

Summary/Abstract: Observations from mixtures of different subpopulations are common in biologicaland sociological studies. We consider the case, when the observations are takenfrom a set of groups containing subjects, which belong to differentsubpopulations. Proportion of each subpopulation in a group is known and canvary from group to group. Our aim is to estimate the means of an observedvariable for subjects, which belong to each subpopulation. In this paper weconsider the case, when subpopulations are defined by answers on so called“sensitive questions”. We consider some parametric and nonparametric estimatesof the subpopulation means, such as weighted means, maximum likelihood andweighted least squares estimates. Finite sample properties of these estimates areanalyzed. Mean square errors of the estimates are compared on simulated data.Some asymptotic results are also given.

• Details
• Contents