Types Of Missing Data Research Articles

Building Cross-National, Longitudinal Data Sets: Issues and Strategies for Implementation

With recent advances in computing power enhancing the ability of social scientists to analyze newly available “big data,” there is a new set of challenges for researchers building cross-national, longitudinal data sets. While scholars should embrace these new sources of data, we must carefully consider cleaning and coding decisions and how these influence our ultimate findings. In this article, we outline five common issues that researchers may face in building large cross-national, longitudinal data sets and suggest strategies for how to address each: (1) country data consistency including births, deaths, splits, unifications, and name changes, (2) longitudinal string matching, (3) identifying different types of missing data, (4) using these types of missing data in developing a theoretically and empirically grounded imputation strategy, and (5) understanding whether systemic change is driven by real world processes or by coding/cleaning choices. We also touch briefly on some general technical and technological considerations when working with large data sets. Throughout, we illustrate issues and strategies with examples drawn from our experience building a cross-national, longitudinal network data set of country-international nongovernmental organization memberships.

Do Participants With Different Patterns of Loss to Follow-Up Have Different Characteristics? A Multi-Wave Longitudinal Study.

BackgroundTo identify patterns of loss to follow-up and baseline predictors of each pattern.MethodsThe Mater-University Study of Pregnancy collected baseline information for 7718 pregnant women who attended Mater Hospital in Brisbane, Australia, from 1981 through 1983. Follow-up data for 6753 eligible participants were collected at 6 months, 5 years, 14 years, 21 years, and 27 years after giving birth. Participants were partitioned into groups of ‘Always Responders’, ‘Returners’, ‘Leavers’, ‘Intermittents’, and ‘Never Responders’. Multinomial logistic regression was used to simultaneously compare baseline characteristics of the last four groups with ‘Always Responders’.ResultsBeing younger, less educated, having no partner, and living in rented housing were associated with being a ‘Returner’. Not owning housing, receiving welfare benefits, and being younger, less educated, not married, a smoker, an Aboriginal/Islander, and born in a non-English-speaking country were associated with being a ‘Leaver’, an ‘Intermittent’, or a ‘Never-responder’. Having higher mental health score and drinking before pregnancy were associated with being a ‘Leaver’ or an ‘Intermittent’. Being unemployed and not physically active were associated with being a ‘Leaver’ or ‘Never Responder’. The groups ‘Leavers’ and ‘Never Responders’ were the most different from the ‘Always Responders’. The group that was most similar to ‘Always Responders’ was the ‘Returners’.ConclusionsPatterns of loss to follow-up should be considered in the application of missing data techniques, where researchers make assumptions about the characteristics of those subjects who do not respond to assess the type of missing data. This information can be used to prevent individuals who are at high risk of dropping out of a study from doing so.

A Sample Selection Model with Skew‐normal Distribution

AbstractNon‐random sampling is a source of bias in empirical research. It is common for the outcomes of interest (e.g. wage distribution) to be skewed in the source population. Sometimes, the outcomes are further subjected to sample selection, which is a type of missing data, resulting in partial observability. Thus, methods based on complete cases for skew data are inadequate for the analysis of such data and a general sample selection model is required. Heckman proposed a full maximum likelihood estimation method under the normality assumption for sample selection problems, and parametric and non‐parametric extensions have been proposed. We generalize Heckman selection model to allow for underlying skew‐normal distributions. Finite‐sample performance of the maximum likelihood estimator of the model is studied via simulation. Applications illustrate the strength of the model in capturing spurious skewness in bounded scores, and in modelling data where logarithm transformation could not mitigate the effect of inherent skewness in the outcome variable.

Analysis of incomplete and inconsistent clinical survey data

It is common for clinical data in survey trials to be incomplete and inconsistent for several reasons. Inconsistent data occur when more than one set of exclusive alternative questions are answered. One objective of this study was to identify and eliminate inconsistent data as an important data mining preprocessing step. We define three types of incomplete data: missing data due to skip pattern (SPMD), undetermined missing data (UMD), and genuine missing data (GMD). Identifying the type of missing data is another important objective as all missing data types cannot be treated the same. This goal cannot be achieved manually on large data of complex surveys since each subject should be processed individually. The analyses are accomplished in a mathematical framework by exploiting graph theoretic structure inherent in the questionnaire. An undirected graph is built using mutually inconsistent responses as well as its complement. The responses not in the largest maximal clique of complement graph are considered inconsistent. This guarantees removing as few responses as possible so that remaining ones are mutually consistent. Further, all potential paths in questionnaire's graph are considered, based on the responses of subjects, to identify each type of incomplete data. Experiments are conducted on MESA data. Results show 15.4 % GMD, 9.8 % SPMD, 12.9 % UMD, and 0.021 % inconsistent data. Further utility of the approach is using a) the SPMD for data stratification, and b) inconsistent data for noise estimation. Proposed method is a preprocessing prerequisite for any data mining of clinical survey data.

Accounting for dropout reason in longitudinal studies with nonignorable dropout

Dropout is a common problem in longitudinal cohort studies and clinical trials, often raising concerns of nonignorable dropout. Selection, frailty, and mixture models have been proposed to account for potentially nonignorable missingness by relating the longitudinal outcome to time of dropout. In addition, many longitudinal studies encounter multiple types of missing data or reasons for dropout, such as loss to follow-up, disease progression, treatment modifications and death. When clinically distinct dropout reasons are present, it may be preferable to control for both dropout reason and time to gain additional clinical insights. This may be especially interesting when the dropout reason and dropout times differ by the primary exposure variable. We extend a semi-parametric varying-coefficient method for nonignorable dropout to accommodate dropout reason. We apply our method to untreated HIV-infected subjects recruited to the Acute Infection and Early Disease Research Program HIV cohort and compare longitudinal CD4+ T cell count in injection drug users to nonusers with two dropout reasons: anti-retroviral treatment initiation and loss to follow-up.

Data Missing Mechanism and Missing Data Real-Time Processing Methods in the Construction Monitoring of Steel Structures

The data link of construction monitoring of steel structures is usually long, complicated and with huge amount of data. Because of the influences of various factors during construction process, sometimes it is inevitable to cause partial monitoring data missing or interrupt of the data link which may affect the real-time control of construction process, even the structural safety assessment during construction. Based on the generating mechanism and characteristics of the missing data discovered from the statistic of actual construction monitoring of steel structures, this paper classifies the missing data into three types for developing new data processing methods. According to the characteristics of the different types of missing data, several compensating methods of the missing data are proposed respectively in this paper by regression analysis based on the principle of least square. The proposed methods are adopted to process the missing monitoring data of the steel structure construction of Art Gallery of Datong and the Circle of Life of Shenyang for verification. The estimated values are compared with the measured data. The numerical results show that the proposed missing data processing methods are effective and practical.

Local Feature Extraction Models from Incomplete Data in Face Recognition Based on Nonnegative Matrix Factorization

Data missing usually happens in the process of data collection, transmission, processing, preservation and application due to various reasons. In the research of face recognition, the missing of image pixel value will affect feature extraction. How to extract local feature from the incomplete data is an interesting as well as important problem. Nonnegative matrix factorization (NMF) is a low rank factorization method for matrix and has been successfully used in local feature extraction in various disciplines which face recognition is included. This paper mainly deals with this problem. Firstly, we classify the patterns of image pixel value missing, secondly, we provide the local feature extraction models basing on nonnegative matrix factorization under different types of missing data, thirdly, we compare the local feature extraction capabilities of the above given models under different missing ratio of the original data. Recognition rate is investigated under different data missing pattern. Numerical experiments are presented and conclusions are drawn at the end of the paper.

Ontological Considerations When Modeling Missing Data With Relational Databases

SQL’s use of nulls to indicate missing data in relational databases has been criticized as violating the relational model. In this article, I review this critique and two popular recommendations of resolution. The first is to retire SQL’s multivalued logic and replace it with a conventional binary logic. The second is to revise SQL’s implementation of multivalued logic so that it can recognize tautological propositions. I argue that underlying this debate is an ontological disagreement about how to properly model missing data and, more generally, social reality. I demonstrate that the relational model provides useful tools for modeling different types of missing data in different ways and, furthermore, offers a useful foundation upon which to conduct social research, one that supports both variable-oriented and case-oriented analysis.

Comparative methods for handling missing data in large databases

Analysis of complex survey databases is an important tool for health services researchers. Missing data elements are challenging because the reasons for "missingness" are multifactorial, especially categorical variables such as race. We simulated missing data for race and analyzed the bias from five methods used in predicting major amputation in patients with critical limb ischemia (CLI). Patient discharges with fully observed data containing lower extremity revascularization or major amputation and CLI were selected from the 2003 to 2007 Nationwide Inpatient Sample, a complex survey database (weighted n = 684,057). Considering several random missing data schemes, we compared five missing data methods: complete case analysis, replacement with observed frequencies, missing indicator variable, multiple imputation, and reweighted estimating equations. We created 100 simulated data sets, with 5%, 15%, or 30% of subjects' race drawn to be missing from the full data set. Bias was estimated by comparing the estimated regression coefficients averaged over 100 simulated data sets (β(miss)) from each method vs estimates from the fully observed data set (β(full)), with relative bias calculated as (β(full) - β(miss)/β(full)) × 100%. Our results demonstrate that reweighted estimating equations produce the least biased and the missing indicator variable produces the most biased coefficients. Complete case analysis, replacement with observed frequencies, and multiple imputation resulted in moderate bias. Sensitivity analysis demonstrated the optimal method choice depends on the quantity and type of missing data encountered. Missing data are an important analytic topic in research with large databases. The commonly used missing indicator variable method introduces severe bias and should be used with caution. We present empiric evidence to guide method selection for handling missing data.

Multiple imputation of missing values was not necessary before performing a longitudinal mixed-model analysis

Background and ObjectivesAs a result of the development of sophisticated techniques, such as multiple imputation, the interest in handling missing data in longitudinal studies has increased enormously in past years. Within the field of longitudinal data analysis, there is a current debate on whether it is necessary to use multiple imputations before performing a mixed-model analysis to analyze the longitudinal data. In the current study this necessity is evaluated. Study Design and SettingThe results of mixed-model analyses with and without multiple imputation were compared with each other. Four data sets with missing values were created—one data set with missing completely at random, two data sets with missing at random, and one data set with missing not at random). In all data sets, the relationship between a continuous outcome variable and two different covariates were analyzed: a time-independent dichotomous covariate and a time-dependent continuous covariate. ResultsAlthough for all types of missing data, the results of the mixed-model analysis with or without multiple imputations were slightly different, they were not in favor of one of the two approaches. In addition, repeating the multiple imputations 100 times showed that the results of the mixed-model analysis with multiple imputation were quite unstable. ConclusionIt is not necessary to handle missing data using multiple imputations before performing a mixed-model analysis on longitudinal data.

Multiple imputation: dealing with missing data

In many fields, including the field of nephrology, missing data are unfortunately an unavoidable problem in clinical/epidemiological research. The most common methods for dealing with missing data are complete case analysis-excluding patients with missing data--mean substitution--replacing missing values of a variable with the average of known values for that variable-and last observation carried forward. However, these methods have severe drawbacks potentially resulting in biased estimates and/or standard errors. In recent years, a new method has arisen for dealing with missing data called multiple imputation. This method predicts missing values based on other data present in the same patient. This procedure is repeated several times, resulting in multiple imputed data sets. Thereafter, estimates and standard errors are calculated in each imputation set and pooled into one overall estimate and standard error. The main advantage of this method is that missing data uncertainty is taken into account. Another advantage is that the method of multiple imputation gives unbiased results when data are missing at random, which is the most common type of missing data in clinical practice, whereas conventional methods do not. However, the method of multiple imputation has scarcely been used in medical literature. We, therefore, encourage authors to do so in the future when possible.

Survival analysis with incomplete genetic data

Genetic data are now collected frequently in clinical studies and epidemiological cohort studies. For a large study, it may be prohibitively expensive to genotype all study subjects, especially with the next-generation sequencing technology. Two-phase sampling, such as case-cohort and nested case-control sampling, is cost-effective in such settings but entails considerable analysis challenges, especially if efficient estimators are desired. Another type of missing data arises when the investigators are interested in the haplotypes or the genetic markers that are not on the genotyping platform used for the current study. Valid and efficient analysis of such missing data is also interesting and challenging. This article provides an overview of these issues and outlines some directions for future research.

The prevention and handling of the missing data

Even in a well-designed and controlled study, missing data occurs in almost all research. Missing data can reduce the statistical power of a study and can produce biased estimates, leading to invalid conclusions. This manuscript reviews the problems and types of missing data, along with the techniques for handling missing data. The mechanisms by which missing data occurs are illustrated, and the methods for handling the missing data are discussed. The paper concludes with recommendations for the handling of missing data.

Mi??ing Data: Should We C?re?

Across all fields of research, the issue of missing data arises. In most studies, surveys, or experiments, there are instances of nonresponse that need to be appropriately dealt with to conduct reasonable statistical analyses. There are several types of missing data as well as ways to deal with incomplete data sets.

Handling of missing data to improve the mining of large feed databases1

Feed databases often have missing data. Despite their potentially major effect on data analysis (e.g., as a source of biased results and loss of statistical power), database managers and nutrition researchers have paid little attention to missing data. This study evaluated various methods of handling missing data using mining outputs from a database containing data on chemical composition and nutritive value for 18,864 alfalfa samples. A complete reference dataset was obtained comprising the 2,303 cases with no missing data for the attributes CP, crude fiber (CF), NDF, ADF and ADL. This dataset was used to simulate 2 types of missing data (at random and not at random), each with 2 loss intensities (33 and 66%), thus yielding a total of 4 incomplete datasets. Missing data from these datasets were handled using 2 deletion methods and 4 imputation methods, and outputs in terms of the identification and typing of alfalfa (using ANOVA and descriptive statistics) and of correlations between attributes (using regressions) were compared with outputs from the complete dataset. Imputation methods, particularly model-based versions, were found to perform better than deletion methods in terms of maximizing information use and minimizing bias although the extent of differences between methods depended on the type of missing data. The best approximation to the uncertainty value was provided by multiple imputation methods. It was concluded that the choice of the most suitable method for handling missing data depended both on the type of missing data and on the purpose of data analysis.

How to avoid missing data and the problems they pose: design considerations.

One of the most common challenges in biomedical and psychosocial research is missing data, which occurs when respondents refuse to provide answers to sensitive questions and when study subjects are lost to follow-up during the repeated assessments of longitudinal trials. This paper is the first in a 3-part series focusing on this important topic; it describes different types of missing data and their differential effects on model estimates, focusing on study design strategies that can be used to prevent or minimize missing data and, thus, maintain the scientific integrity of the research. The second paper in the series will discuss implementation strategies to manage and reduce missing data while conducting the study, and the third paper will discuss analytic strategies for dealing with missing data after completion of data collection. It is always worth devoting careful attention to issues like missing data that may have significant effects on model estimates. As the saying goes, ‘an ounce of prevention is worth a pound of cure,’ so it is much better to focus on these issues during the planning stage of a study rather than having to deal with them later in the study. In this paper, we focus squarely on such preventive strategies as the first line of defense against the ubiquitous problem of missing data in clinical research studies. 1. Types of missing data and their effects on model estimates The reasons for missing data vary, and the degree to which missing data decreases the validity of the estimates depends on how the missing data arises. Thus it is important to make plausible assumptions about how missing data occurs in a study and, based on these assumptions, select appropriate models for addressing the effect of the missing data on inference from the observed results of the study. There are three statistical models with increasing levels of generalizability that are commonly used to classify different types of missing data. If missing data occur in a random fashion—that is, with no particular pattern that determines which data are observed and which are missing—it is typically referred to as ‘missing completely at random’ (MCAR). Data that are MCAR have no influence on any of the study participants’ outcomes and, thus, may be ignored because they do not result in an inferential bias. In many follow-up studies participants may be lost to follow up because of deteriorated or improved health conditions. In this situation the missed data are not MCAR because the probability of the missed visit depends on the outcome. For example, if an investigational medication worsens depression, subjects may drop out from the study over time, creating a so-called ‘monotone missing data’ pattern. In cases like this where whether or not data is missing is influenced by treatment-related effects, ignoring the missing data and focusing only on the subjects with complete data will usually give rise to biased estimates of treatment effects. If depression measures taken at baseline (or prior to dropping out) can be used to model this dependence structure in the missing data pattern (i.e., the relationship between baseline severity of depression and dropping out) this information can then be incorporated into the model for treatment effects to address the bias. If the dependence structure between the outcome of interest and the missing data can be modeled based on observed data, the missing data are classified as ‘missing at random’ (MAR).

Longitudinal data analysis with non-ignorable missing data

A common problem in the longitudinal data analysis is the missing data problem. Two types of missing patterns are generally considered in statistical literature: monotone and non-monotone missing data. Nonmonotone missing data occur when study participants intermittently miss scheduled visits, while monotone missing data can be from discontinued participation, loss to follow-up, and mortality. Although many novel statistical approaches have been developed to handle missing data in recent years, few methods are available to provide inferences to handle both types of missing data simultaneously. In this article, a latent random effects model is proposed to analyze longitudinal outcomes with both monotone and non-monotone missingness in the context of missing not at random. Another significant contribution of this article is to propose a new computational algorithm for latent random effects models. To reduce the computational burden of high-dimensional integration problem in latent random effects models, we develop a new computational algorithm that uses a new adaptive quadrature approach in conjunction with the Taylor series approximation for the likelihood function to simplify the E-step computation in the expectation-maximization algorithm. Simulation study is performed and the data from the scleroderma lung study are used to demonstrate the effectiveness of this method.

Reweighted Mahalanobis distance matching for cluster‐randomized trials with missing data

This paper introduces an improved tool for designing matched-pairs randomized trials. The tool allows the incorporation of clinical and other knowledge regarding the relative importance of variables used in matching and allows for multiple types of missing data. The method is illustrated in the context of a cluster-randomized trial. A Web application and an R package are introduced to implement the method and incorporate recent advances in the area. Reweighted Mahalanobis distance (RMD) matching incorporates user-specified weights and imputed values for missing data. Weight may be assigned to missingness indicators to match on missingness patterns. Three examples are presented, using real data from a cohort of 90 Veterans Health Administration sites that had at least 100 incident metformin users in 2007. Matching is utilized to balance seven factors aggregated at the site level. Covariate balance is assessed for 10,000 randomizations under each strategy: simple randomization, matched randomization using the Mahalanobis distance, and matched randomization using the RMD. The RMD matching achieved better balance than simple randomization or MD randomization. In the first example, simple and MD randomization resulted in a 10% chance of seeing an absolute mean difference of greater than 26% in the percent of nonwhite patients per site; the RMD dramatically reduced that to 6%. The RMD achieved significant improvement over simple randomization even with as much as 20% of the data missing. Reweighted Mahalanobis distance matching provides an easy-to-use tool that incorporates user knowledge and missing data.

Effect of different statistical methods on union or time to union in a published study about clavicular fractures

Time to union is a suspect measure for comparing treatments given the absence of a consensus definition of union, the limited reliability of diagnostic tests, and inconsistency in evaluation times. The purpose of this study was to quantify the variations in union and time to union according to different statistical methods and different approaches to missing data. Data from a published multicenter, randomized trial comparing operative and nonoperative treatment of clavicular fractures were reanalyzed. Two main types of missing data were encountered: (1) lost to follow-up or died before union and (2) missed appointment. We studied the effect of four statistical methods-comparison of means, comparison of medians, χ(2), and Kaplan-Meier curves-for comparing union or time to union between cohorts for the following scenarios: strict intention-to-treat, intention-to-treat with exclusion of patients with less than 12 months of follow-up, as-treated analysis, and four different imputation methods for missing data. Mean and median time to union varied up to 17%, but comparative statistics consistently demonstrated shorter time to union among operatively treated patients. There were significant differences in the odds ratio, χ(2) values, and the number needed to treat (8%-62%) of union vs nonunion for the three principal analyses. Different strategies for handling missed evaluations seem to influence categoric results (eg, union or nonunion) more than continuous measures such as time to union.

Data Envelopment Analysis of clinics with sparse data: Fuzzy clustering approach

This paper presents a method for utilizing Data Envelopment Analysis (DEA) with sparse input and output data using fuzzy clustering concepts. DEA, a methodology to assess relative technical efficiency of production units is susceptible to missing data, thus, creating a need to supplement sparse data in a reliable and accurate manner. The approach presented is based on a modified fuzzy c-means clustering using optimal completion strategy (OCS) algorithm. This particular algorithm is sensitive to the initial values chosen to substitute missing values and also to the selected number of clusters. Therefore, this paper proposes an approach to estimate the missing values using the OCS algorithm, while considering the issue of initial values and cluster size. This approach is demonstrated on a real and complete dataset of 22 rural clinics in the State of Kansas, assuming varying levels of missing data. Results show the effect of the clustering based approach on the data recovered considering the amount and type of missing data. Moreover, the paper shows the effect that the recovered data has on the DEA scores.