Comparison of Tests and Confidence Intervals for Univariate Normal Mean Based on Multiply Imputed Synthetic Data Obtained by Posterior Predictive Sampling

Abstract

There is a huge literature on data analysis under privacy or confidentiality protection. Among many inferential statistical methods based on parametric models, data analysis based on perturbation of original sensitive data using plug-in and posterior predictive sampling are quite common. In this paper we consider a very basic inferential problem of tests and confidence intervals for a normal mean with unknown variance based on synthetic data obtained from multiple imputations under posterior predictive sampling method. Several methods are suggested and compared. A general expression of the local power of a class of tests is also derived which can be used in a design context to determine a combination of sample size and number of imputations to guarantee a desired level of local power. A measure of privacy protection is derived to demonstrate that privacy would be compromised if too many imputations are released. An application to draw inference about the household earnings, corresponding to a US Census Bureau data, is illustrated.

Keywords

Local power multiple imputation normal mean posterior predictive sampling synthetic data privacy protection

Get full access to this article

View all access options for this article.

References

Dalenius

and Reiss

. Data-swapping: A technique for disclosure control. J Stat Plan Infer 1982; 6(1): 73–85.

Nayak

, Sinha

and Zayatz

Statistical properties of multiplicative noise masking for confidentiality protection. J Off Stat 2011 ; 27(3): 527–544.

Sinha

, Nayak

and Zayatz

Privacy protection and quantile estimation from noise multiplied data. Sankhya B 2011; 73: 297–315.

Klein

and Sinha

Statistical analysis of noise multiplied data using multiple imputation. J Off Stat 2013; 29(3): 425–465.

Klein

, Mathew

and Sinha

Noise multiplication for statistical disclosure control of extreme values in log-normal regression samples. J Priv Confidentiality 2014; 6(1): 77–125.

Reiter

JP.

Inference for partially synthetic, public use microdata sets. Surv Methodol 2003; 29(2): 181–188.

Reiter

JP.

Releasing multiply imputed, synthetic public use microdata: an illustration and empirical study. J R Stat Soc Ser A: Stat Soc 2005; 168(1): 185–205.

Reiter

JP.

Significance tests for multi-component estimands from multiply imputed, synthetic microdata. J Stat Plan Infer 2005; 131(2): 365–377.

Klein

and Sinha

Likelihood-based inference for singly and multiply imputed synthetic data under a normal model. Stat Probab Lett 2015; 105: 168–175.

10.

Klein

and Sinha

Inference for singly imputed synthetic data based on posterior predictive sampling under multivariate normal and multiple linear regression models. Sankhya B 2015; 77(2): 293–311.

11.

Klein

and Sinha

Likelihood-based finite sample inference for singly imputed synthetic data under multivariate normal and multiple linear regression models. J Priv and Confidentiality 2016; 7(1): 43–98.

12.

Moura

and Klein

and Coelho

, . Inference for multivariate regression model based on synthetic data generated under posterior and fixed posterior predictive sampling: comparison with plug-in sampling. Revstat - Stat J 2016; 15(2).

13.

Klein

and Sinha

Multiple imputation for parametric inference under a differentially private Laplace mechanism . Technical Report, Department of Math and Statistics, UMBC, 2019.

14.

Klein

, Zylstra

and Sinha

Finite sample inference for multiply imputed synthetic data under a multiple linear regression model. Calcutta Stat Assoc Bull 2019; 71(2): 63–82.

15.

Klein

, Moura

, & Sinha

Multivariate normal inference based on singly imputed synthetic data under plug-in sampling. Sankhya B 2021;83-B(Part 1): 273–287.

16.

Moura

, Klein

, Zylstra

, . Inference for multivari- ate regression model based on synthetic data generated using plug-in sampling. J Am Stat Assoc 2021; 116(534): 1–41.

17.

Guin

, Roy

and Sinha

Bayesian analysis of singly imputed partially synthetic data generated by plug-in sampling and posterior predictive sampling under the multiple linear regression model. Int J Stat Sci 2021.

18.

Guin

, Roy

and Sinha

Bayesian analysis of multiply imputed synthetic data under the multiple linear regression model. Int J Stat Sci 2022; 25–38.

19.

Guin

, Roy

and Sinha

Bayesian analysis of singly imputed synthetic data under the multivariate normal model. Int J Stat Sci 2023; 23(2): 1–18, 2023.

20.

Hartung

, Knapp

and Sinha

BK.

Statistical meta-analysis with applications . New Jersey: Wiley, 2008.

21.

Graybill

, & Deal

RB.

Combining unbiased estimators. Biometrics 1959; 15(4): 543–550.

22.

Khatri

, & Shah

KR.

Estimation of location parameters from two linear models under normality. Commun Stat 1974; 3(7): 647–663.

23.

Meier

Variance of a weighted mean. Biometrics 1953; 9(1): 59–73.

24.

Sinha

BK.

Unbiased estimation of the variance of the graybill-deal estimator of the common mean of several normal populations. Can J Stat 1985; 13(3): 243–247.

25.

Tippett

LHC.

The methods of statistics . London: Williams and Norgate Ltd., 1931.