Some Imputation Methods to Minimize the Effect of Non Response in Two-Occasion Rotation Patterns

Among modern strategies applied to cope with non response, a major problem faced by survey statisticians, imputation is one of the most common. Imputation is the filling up method of incomplete data for adapting the standard analytic model in statistics. The purpose of the present work is to study the use of imputation methods in dealing with non response which occur at both occasions in two-occasion successive (rotation) sampling. Chain-type regressions in ratio estimators have been proposed for estimating the population mean at current occasion. Expressions for optimum estimator and its mean square error have been derived. To study the effectiveness of the imputation methods, performances of the proposed estimators are compared in two different situations: with and without non response. Behavior of the proposed estimators is demonstrated through empirical studies.     

Effectual variance estimation strategy in two-occasion successive sampling in presence of random non response

Present investigation deals with the problem of random non response in estimation of population variance on current occasion in two-occasion successive sampling. To reduce the negative impact of random non response on both occasions, regression type imputation method has been suggested. Using auxiliary information, efficient estimation strategies have been developed for estimation of population variance on the current occasion. Estimator for the population variance is also derived as a special case when random non response occurs only on the current occasion. Empirical studies are carried out to show the dominance of suggested estimators over sample variance and exponential type estimators. Results are interpreted.     

Estimation of population mean in successive sampling under super-population model in presence of random non response situations

The present work is an attempt to estimate the population mean on the current occasion in two-occasion successive (rotation) sampling in presence of random non response situations. The estimation strategy has been constructed under a super-population model design approach with the help of imputation technique. The estimators proposed on the current occasion cover the cases of occurrences random non responses on either of the occasions. Detail behaviors of the proposed class of estimators have been studied and its performance has been examined with the sample mean estimator. The results are demonstrated through empirical studies which establish the effectiveness of the proposed class of estimators. Suitable recommendations have been put forward to the survey statisticians for its practical application.     

Use of multiauxiliary variables to cope with non response in estimation of current population mean in two-occasion successive sampling

ABSTRACT

The present work utilizes the information on multiauxiliary variables to neutralize the negative effect of non response in estimation of current population mean in two-occasion successive sampling. Generalized exponential type estimators have been proposed for the situation when non response occurs at both occasion or at first occasion or at current (second) occasion in two-occasion successive sampling. Properties of the proposed estimators have studied and empirical studies are carried out to justify the preposition of estimators. Suitable recommendations have been made.     

Estimation of Population Mean Under Non Response in Two-Occasion Rotation Patterns

The present article is an attempt to study the effect of non response at both occasions in search of good rotation patterns over two occasions. Ratio-type estimators were proposed for estimating the population mean at current occasion in presence of non response at both the occasions in two-occasion successive (rotation) sampling. Detail behaviors of proposed estimators were studied. Proposed estimators were compared with the estimator using no information from previous (first) occasion. Performances of the proposed estimators were demonstrated via empirical studies.     

Effective Rotation Patterns Under Non Response in Two-Occasion Successive Sampling

This article intends to develop some effective rotation patterns with the aid of attractive imputation methods when the problems of non response occur in two-occasion successive sampling. Utilizing the information on p (p ??1) auxiliary variables regression methods of imputation have been considered and subsequently multiple linear regression type estimators are proposed to estimate the current population mean in two-occasion successive sampling. Proposed estimators are compared with the estimator for same situations but in the absence of non-response. Optimum replacement strategies of the respective estimators have been discussed and results are interpreted with the help of empirical studies. Conclusions and suitable recommendations are made.     

Some efficient classes of estimators of population mean in two-phase successive sampling under random non response

In this paper, an attempt has been made to present some efficient classes of estimators of population mean on current occasion in two-occasion successive sampling under random non response in two-phase setup. Effectiveness of the proposed classes of estimators has been studied under the assumptions that sampling units follow a distribution under the random non response. To check the performances, the proposed classes of estimators are compared with an estimator under the similar situation under the complete response. Results are demonstrated through empirical studies which show the reliable nature of the proposed classes of estimators. Suitable recommendations have been made to the survey practitioners.     

Some effective combinations of available information under non response in two-occasion successive sampling

The present work suggests some effective linear combinations of exponential estimators to estimate the current population mean in two occasion successive sampling in presence of non response. Using the subsampling technique of non respondents, estimation procedures have been developed by utilizing information on an auxiliary variable, which is readily available on both occasions. Properties of the proposed estimation procedures have been examined and related optimum replacement strategies are suggested. Empirical studies are carried out to justify the preposition of estimators. Suitable recommendations have been made.     

Doubly Robust Inference for the Distribution Function in the Presence of Missing Survey Data

Item non‐response in surveys occurs when some, but not all, variables are missing. Unadjusted estimators tend to exhibit some bias, called the non‐response bias, if the respondents differ from the non‐respondents with respect to the study variables. In this paper, we focus on item non‐response, which is usually treated by some form of single imputation. We examine the properties of doubly robust imputation procedures, which are those that lead to an estimator that remains consistent if either the outcome variable or the non‐response mechanism is adequately modelled. We establish the double robustness property of the imputed estimator of the finite population distribution function under random hot‐deck imputation within classes. We also discuss the links between our approach and that of Chambers and Dunstan. The results of a simulation study support our findings.     

On the use of multi-auxiliary variables to neutralize the effect of non response in two-occasion successive sampling

In this work, we have suggested some linear combinations of generalized ratio type estimators in estimation of current population mean when non response occur in two occasion successive sampling. Information on multi-auxiliary variables has been used and sub-sampling of non respondents technique is utilized to cope with the negative effect of non response. Properties of the suggested estimation procedures have been examined and suitable optimum replacement strategies are discussed. Empirical studies are carried out to justify the proposition of estimators. Suitable recommendations have been made.     

Application of an imputation method for variance estimation under pseudo-likelihood when missing data are NMAR

When data are outcome-dependent non response, pseudo-likelihood yields consistent regression coefficients without specifying the missing data mechanism. However, it is onerous to derive parameter estimators including their standard errors from the regression coefficients under pseudo-likelihood (PL). The present study applies an imputation method to compute the asymptotic standard errors of parameter estimators. The proposed method is simpler than Delta method and it showed similar effect size of the standard errors to bootstrapping in simulation and application studies.     

Estimation of population mean in presence of random non response in two-stage cluster sampling

The present investigation addresses the problem of estimating a finite population mean in two-phase cluster sampling in presence of random non response situations. Utilizing information on an auxiliary variable, regression type estimators has been proposed. Effective imputation techniques have been suggested to deal with the random non response situations. The properties of the proposed estimation strategies have been studied for different cases of random non response situations in practical surveys. The superiority of the suggested methodology over the natural sample mean estimator of population mean has been established through empirical studies carried over the data sets of natural population and artificially generated population.     

Estimation of population mean on the most recent occasion under non response in h-occasion successive sampling using several auxiliary variables

In this article, general estimation procedures of population mean on the most recent occasion in the presence of non response on all occasions have been deduced in h-occasion successive sampling. In the estimation procedures, regression type estimators have been suggested with the aid of stable and mutually independent several auxiliary variables which are readily available only on the most recent occasion. To check the efficiency of the proposed estimators, they have been compared with similar estimator in absence of non response. Properties of the suggested estimation procedures have been studied and their empirical studies are carried out to validate the theoretical results. Suitable recommendations have been made.     

Parametric fractional imputation for nonignorable missing data

Parameter estimation with missing data is a frequently encountered problem in statistics. Imputation is often used to facilitate the parameter estimation by simply applying the complete-sample estimators to the imputed dataset.In this article, we consider the problem of parameter estimation with nonignorable missing data using the approach of parametric fractional imputation proposed by Kim (2011). Using the fractional weights, the E-step of the EM algorithm can be approximated by the weighted mean of the imputed data likelihood where the fractional weights are computed from the current value of the parameter estimates. Calibration fractional imputation is also considered as a way for improving the Monte Carlo approximation in the fractional imputation. Variance estimation is also discussed. Results from two simulation studies are presented to compare the proposed method with the existing methods. A real data example from the Korea Labor and Income Panel Survey (KLIPS) is also presented.     

Empirical likelihood confidence intervals under imputation for missing survey data from stratified simple random sampling

Missing observations due to non‐response are commonly encountered in data collected from sample surveys. The focus of this article is on item non‐response which is often handled by filling in (or imputing) missing values using the observed responses (donors). Random imputation (single or fractional) is used within homogeneous imputation classes that are formed on the basis of categorical auxiliary variables observed on all the sampled units. A uniform response rate within classes is assumed, but that rate is allowed to vary across classes. We construct confidence intervals (CIs) for a population parameter that is defined as the solution to a smooth estimating equation with data collected using stratified simple random sampling. The imputation classes are assumed to be formed across strata. Fractional imputation with a fixed number of random draws is used to obtain an imputed estimating function. An empirical likelihood inference method under the fractional imputation is proposed and its asymptotic properties are derived. Two asymptotically correct bootstrap methods are developed for constructing the desired CIs. In a simulation study, the proposed bootstrap methods are shown to outperform traditional bootstrap methods and some non‐bootstrap competitors under various simulation settings. The Canadian Journal of Statistics 47: 281–301; 2019 © 2019 Statistical Society of Canada     

Statistical inference under imputation for proportional hazard model with missing covariates

Missing covariate data are common in biomedical studies. In this article, by using the non parametric kernel regression technique, a new imputation approach is developed for the Cox-proportional hazard regression model with missing covariates. This method achieves the same efficiency as the fully augmented weighted estimators (Qi et al. 2005. Journal of the American Statistical Association, 100:1250) and has a simpler form. The asymptotic properties of the proposed estimator are derived and analyzed. The comparisons between the proposed imputation method and several other existing methods are conducted via a number of simulation studies and a mouse leukemia data.     

Empirical likelihood for generalized linear models with missing responses

The paper uses the empirical likelihood method to study the construction of confidence intervals and regions for regression coefficients and response mean in generalized linear models with missing response. By using the inverse selection probability weighted imputation technique, the proposed empirical likelihood ratios are asymptotically chi-squared. Our approach is to directly calibrate the empirical likelihood ratio, which is called as a bias-correction method. Also, a class of estimators for the parameters of interest is constructed, and the asymptotic distributions of the proposed estimators are obtained. A simulation study indicates that the proposed methods are comparable in terms of coverage probabilities and average lengths/areas of confidence intervals/regions. An example of a real data set is used for illustrating our methods.     

A new method of imputation in survey sampling

In this paper, an alternative estimator of population mean in the presence of non-response has been suggested which comes in the form of Walsh's estimator. The estimator of mean obtained from the proposed technique remains better than the estimators obtained from ratio or mean methods of imputation. The mean-squared error (MSE) of the resultant estimator is less than that of the estimator obtained on the basis of ratio method of imputation for the optimum choice of parameters. An estimator for estimating a parameter involved in the process of new method of imputation has been discussed. A suggestion to form ‘warm deck’ method of imputation has been suggested. The MSE expressions for the proposed estimators have been derived analytically and compared empirically. The work has been extended to the case of multi-auxiliary information to be used for imputation. Numerical illustrations are also provided.     

Search of Good Rotation Patterns to Improve the Precision of Estimates at Current Occasion

In rotation (successive) sampling, it is common practice to use the information collected on a previous occasion to improve the precision of the estimates at current occasion. The previous information may be in the form of an auxiliary character, the character under study itself, or both. In the present work, information on an auxiliary character, which is readily available on all the occasions, has been used along with the information on study character from the previous and current occasion. Consequently, chain type difference and regression estimators have been proposed for estimating the population mean at second (current) occasion in the two occasions rotation (successive) sampling. The proposed estimators have been compared with sample mean estimator when there is no matching and the optimum estimator, which is the combination of the means of the matched and unmatched portions of the sample at the second occasion. Optimum replacement policy is also discussed. Theoretical results have been justified through empirical interpretation.     

Imputation by power transformation

In this paper, a new power transformation estimator of population mean in the presence of non-response has been suggested. The estimator of mean obtained from proposed technique remains better than the estimators obtained from ratio or mean methods of imputation. The mean squared error of the resultant estimator is less than that of the estimator obtained on the basis of ratio method of imputation for the optinum choice of parameters. An estimator for estimating a parameter involved in the process of new method of imputation has been discussed. The MSE expressions for the proposed estimators have been derived analytically and compared empirically. Product method of imputation for negatively correlated variables has also been introduced. The work has been extended to the case of multi-auxiliary information to be used for imputation.