Empirical Likelihood Methods for Complex Surveys With Data Missing-by-Design

Abstract

We consider nonrandomized pretest-posttest designs with complex survey data for observational studies. We show that two-sample pseudo empirical likelihood methods provide efficient inferences on the treatment effect, with a missing-by-design feature used for forming the two samples and the baseline information incorporated through suitable constraints. The proposed maximum pseudo empirical likelihood estimators of the treatment effect are consistent and pseudo empirical likelihood ratio confidence intervals are constructed through bootstrap calibration methods. The proposed methods require estimation of propensity scores which depend on the underlying missing-by-design mechanism.A simulation study is conducted to examine finite sample performances of the proposed methods under different scenarios of nonignorable and ignorable missing patterns. An application to the International Tobacco Control Policy Evaluation Project Four Country Surveys is also presented to demonstrate the use of the proposed methods for examining the mode effect in survey data collection.

Calibration Techniques Encompassing Survey Sampling, Missing Data Analysis and Causal Inference

“…The constraint (8) can be interpreted as matching the calibration estimator ∑ i ∈ S p i μðx i ; θÞ to the corresponding population mean μ N ¼ N À1 ∑ N i¼1 μðx i ; θÞ. The ℓ PEL ðpÞ given in ( 6) is termed as the pseudo empirical log-likelihood function by Chen & Sitter (1999) because it is a survey-weighted version of the following empirical log-likelihood function proposed by Owen (1988) for i.i.d. sample data:…”

Combining Non-Probability and Probability Survey Samples Through Mass Imputation

“…There have been reported survey mode effect in the literature that measurements on behaviours such as smoking and drinking differ significantly between web surveys and telephone surveys. See, for instance, Chen et al (2018) for findings from the International Tobacco Control (ITC) Canada survey. The Hájek estimator in the current analysis provides an estimate for the population means of Z 1 and Z 2 with measurements taken by telephone interviews.…”

“…See, for instance, Chen et al. (2018) for findings from the International Tobacco Control (ITC) Canada survey. The Hájek estimator in the current analysis provides an estimate for the population means of and with measurements taken by telephone interviews.…”

Recent Developments in Dealing with Item Non‐response in Surveys: A Critical Review

“…Deriving consistent variance estimators for distribution functions and quantiles is much more challenging. Chen et al () developed pseudo‐population bootstrap procedures for estimating the variance of distribution functions and quantiles. The construction of the pseudo‐populations is based on .…”