Multiple Imputation for Continuous and Categorical Data: Comparing Joint Multivariate Normal and Conditional Approaches

2014 ◽  
Vol 22 (4) ◽  
pp. 497-519 ◽  
Author(s):  
Jonathan Kropko ◽  
Ben Goodrich ◽  
Andrew Gelman ◽  
Jennifer Hill
Keyword(s):  
Multiple Imputation ◽  
Categorical Data ◽  
Missing Values ◽  
Missing At Random ◽  
Model Fit ◽  
Categorical Variables ◽  
Data Sets ◽  
Multivariate Normal ◽  
Evaluating Methods ◽  
Election Studies

We consider the relative performance of two common approaches to multiple imputation (MI): joint multivariate normal (MVN) MI, in which the data are modeled as a sample from a joint MVN distribution; and conditional MI, in which each variable is modeled conditionally on all the others. In order to use the multivariate normal distribution, implementations of joint MVN MI typically assume that categories of discrete variables are probabilistically constructed from continuous values. We use simulations to examine the implications of these assumptions. For each approach, we assess (1) the accuracy of the imputed values; and (2) the accuracy of coefficients and fitted values from a model fit to completed data sets. These simulations consider continuous, binary, ordinal, and unordered-categorical variables. One set of simulations uses multivariate normal data, and one set uses data from the 2008 American National Election Studies. We implement a less restrictive approach than is typical when evaluating methods using simulations in the missing data literature: in each case, missing values are generated by carefully following the conditions necessary for missingness to be “missing at random” (MAR). We find that in these situations conditional MI is more accurate than joint MVN MI whenever the data include categorical variables.

scholarly journals Multiple Imputation Approaches Applied to the Missing Value Problem in Bottom-Up Proteomics

2021 ◽  
Vol 22 (17) ◽  
pp. 9650
Author(s):  
Miranda L. Gardner ◽  
Michael A. Freitas
Keyword(s):  
Multiple Imputation ◽  
Missing At Random ◽  
Data Sets ◽  
Proteomics Data ◽  
Missing Not At Random ◽  
Differential Abundance ◽  
Missing Value ◽  
Bottom Up ◽  
Missing Value Imputation ◽  
Impute Data

Analysis of differential abundance in proteomics data sets requires careful application of missing value imputation. Missing abundance values widely vary when performing comparisons across different sample treatments. For example, one would expect a consistent rate of “missing at random” (MAR) across batches of samples and varying rates of “missing not at random” (MNAR) depending on the inherent difference in sample treatments within the study. The missing value imputation strategy must thus be selected that best accounts for both MAR and MNAR simultaneously. Several important issues must be considered when deciding the appropriate missing value imputation strategy: (1) when it is appropriate to impute data; (2) how to choose a method that reflects the combinatorial manner of MAR and MNAR that occurs in an experiment. This paper provides an evaluation of missing value imputation strategies used in proteomics and presents a case for the use of hybrid left-censored missing value imputation approaches that can handle the MNAR problem common to proteomics data.

scholarly journals Comparison of Selected Multiple Imputation Methods for Continuous Variables – Preliminary Simulation Study Results

2019 ◽  
Vol 6 (339) ◽  
pp. 73-98
Author(s):  
Małgorzata Aleksandra Misztal
Keyword(s):  
Missing Data ◽  
Multiple Imputation ◽  
Missing Values ◽  
Imputation Accuracy ◽  
Imputation Method ◽  
Data Sets ◽  
Continuous Variables ◽  
Imputation Methods ◽  
Study Results ◽  
Almost All

The problem of incomplete data and its implications for drawing valid conclusions from statistical analyses is not related to any particular scientific domain, it arises in economics, sociology, education, behavioural sciences or medicine. Almost all standard statistical methods presume that every object has information on every variable to be included in the analysis and the typical approach to missing data is simply to delete them. However, this leads to ineffective and biased analysis results and is not recommended in the literature. The state of the art technique for handling missing data is multiple imputation. In the paper, some selected multiple imputation methods were taken into account. Special attention was paid to using principal components analysis (PCA) as an imputation method. The goal of the study was to assess the quality of PCA‑based imputations as compared to two other multiple imputation techniques: multivariate imputation by chained equations (MICE) and missForest. The comparison was made by artificially simulating different proportions (10–50%) and mechanisms of missing data using 10 complete data sets from the UCI repository of machine learning databases. Then, missing values were imputed with the use of MICE, missForest and the PCA‑based method (MIPCA). The normalised root mean square error (NRMSE) was calculated as a measure of imputation accuracy. On the basis of the conducted analyses, missForest can be recommended as a multiple imputation method providing the lowest rates of imputation errors for all types of missingness. PCA‑based imputation does not perform well in terms of accuracy.

scholarly journals Fitting Ordinal Factor Analysis Models With Missing Data: A Comparison Between Pairwise Deletion and Multiple Imputation

2019 ◽  
Vol 80 (1) ◽  
pp. 41-66 ◽  
Author(s):  
Dexin Shi ◽  
Taehun Lee ◽  
Amanda J. Fairchild ◽  
Alberto Maydeu-Olivares
Keyword(s):  
Factor Analysis ◽  
Missing Data ◽  
Multiple Imputation ◽  
Missing At Random ◽  
Complete Data ◽  
Model Fit ◽  
Parameter Estimates ◽  
Fit Indices ◽  
Wide Range ◽  
Analysis Models

This study compares two missing data procedures in the context of ordinal factor analysis models: pairwise deletion (PD; the default setting in Mplus) and multiple imputation (MI). We examine which procedure demonstrates parameter estimates and model fit indices closer to those of complete data. The performance of PD and MI are compared under a wide range of conditions, including number of response categories, sample size, percent of missingness, and degree of model misfit. Results indicate that both PD and MI yield parameter estimates similar to those from analysis of complete data under conditions where the data are missing completely at random (MCAR). When the data are missing at random (MAR), PD parameter estimates are shown to be severely biased across parameter combinations in the study. When the percentage of missingness is less than 50%, MI yields parameter estimates that are similar to results from complete data. However, the fit indices (i.e., χ2, RMSEA, and WRMR) yield estimates that suggested a worse fit than results observed in complete data. We recommend that applied researchers use MI when fitting ordinal factor models with missing data. We further recommend interpreting model fit based on the TLI and CFI incremental fit indices.

scholarly journals A Note on Listwise Deletion versus Multiple Imputation

2018 ◽  
Vol 26 (4) ◽  
pp. 480-488 ◽  
Author(s):  
Thomas B. Pepinsky
Keyword(s):  
Multiple Imputation ◽  
Missing Values ◽  
Missing At Random ◽  
Strong Correlations ◽  
Simulation Approach ◽  
Missing Not At Random ◽  
Listwise Deletion ◽  
Data Generating Process

This letter compares the performance of multiple imputation and listwise deletion using a simulation approach. The focus is on data that are “missing not at random” (MNAR), in which case both multiple imputation and listwise deletion are known to be biased. In these simulations, multiple imputation yields results that are frequently more biased, less efficient, and with worse coverage than listwise deletion when data are MNAR. This is the case even with very strong correlations between fully observed variables and variables with missing values, such that the data are very nearly “missing at random.” These results recommend caution when comparing the results from multiple imputation and listwise deletion, when the true data generating process is unknown.

NORM software review: Handling missing values with multiple imputation methods

2002 ◽  
Vol 2 (1) ◽  
pp. 51-57 ◽  
Author(s):  
I Gusti Ngurah Darmawan
Keyword(s):  
Missing Data ◽  
Multiple Imputation ◽  
Missing Values ◽  
Evaluation Studies ◽  
Data Uncertainty ◽  
Small Data ◽  
Data Sets ◽  
Computational Statistics ◽  
Adoption And Implementation ◽  
Small Data Sets

Evaluation studies often lack sophistication in their statistical analyses, particularly where there are small data sets or missing data. Until recently, the methods used for analysing incomplete data focused on removing the missing values, either by deleting records with incomplete information or by substituting the missing values with estimated mean scores. These methods, though simple to implement, are problematic. However, recent advances in theoretical and computational statistics have led to more flexible techniques with sound statistical bases. These procedures involve multiple imputation (MI), a technique in which the missing values are replaced by m > 1 estimated values, where m is typically small (e.g. 3-10). Each of the resultant m data sets is then analysed by standard methods, and the results are combined to produce estimates and confidence intervals that incorporate missing data uncertainty. This paper reviews the key ideas of multiple imputation, discusses the currently available software programs relevant to evaluation studies, and demonstrates their use with data from a study of the adoption and implementation of information technology in Bali, Indonesia.

scholarly journals Estimating Average Treatment Effects Utilizing Fractional Imputation when Confounders are Subject to Missingness

2020 ◽  
Vol 8 (1) ◽  
pp. 249-271
Author(s):  
Nathan Corder ◽  
Shu Yang
Keyword(s):  
Multiple Imputation ◽  
Treatment Effects ◽  
Missing Values ◽  
Missing At Random ◽  
Average Treatment Effect ◽  
Asymptotically Normal ◽  
Average Treatment ◽  
Accuracy And Precision ◽  
Causal Treatment ◽  
Average Treatment Effects

Abstract The problem of missingness in observational data is ubiquitous. When the confounders are missing at random, multiple imputation is commonly used; however, the method requires congeniality conditions for valid inferences, which may not be satisfied when estimating average causal treatment effects. Alternatively, fractional imputation, proposed by Kim 2011, has been implemented to handling missing values in regression context. In this article, we develop fractional imputation methods for estimating the average treatment effects with confounders missing at random. We show that the fractional imputation estimator of the average treatment effect is asymptotically normal, which permits a consistent variance estimate. Via simulation study, we compare fractional imputation’s accuracy and precision with that of multiple imputation.

scholarly journals Missing not at random in end of life care studies: multiple imputation and sensitivity analysis on data from the ACTION study

2021 ◽  
Vol 21 (1) ◽  
Author(s):  
Giulia Carreras ◽  
◽  
Guido Miccinesi ◽  
Andrew Wilcock ◽  
Nancy Preston ◽  
...  
Keyword(s):  
Sensitivity Analysis ◽  
Multiple Imputation ◽  
End Of Life ◽  
End Of Life Care ◽  
Missing Values ◽  
Controlled Trial ◽  
Missing At Random ◽  
Life Care ◽  
Missing Not At Random ◽  
Cluster Randomized

Abstract Background Missing data are common in end-of-life care studies, but there is still relatively little exploration of which is the best method to deal with them, and, in particular, if the missing at random (MAR) assumption is valid or missing not at random (MNAR) mechanisms should be assumed. In this paper we investigated this issue through a sensitivity analysis within the ACTION study, a multicenter cluster randomized controlled trial testing advance care planning in patients with advanced lung or colorectal cancer. Methods Multiple imputation procedures under MAR and MNAR assumptions were implemented. Possible violation of the MAR assumption was addressed with reference to variables measuring quality of life and symptoms. The MNAR model assumed that patients with worse health were more likely to have missing questionnaires, making a distinction between single missing items, which were assumed to satisfy the MAR assumption, and missing values due to completely missing questionnaire for which a MNAR mechanism was hypothesized. We explored the sensitivity to possible departures from MAR on gender differences between key indicators and on simple correlations. Results Up to 39% of follow-up data were missing. Results under MAR reflected that missingness was related to poorer health status. Correlations between variables, although very small, changed according to the imputation method, as well as the differences in scores by gender, indicating a certain sensitivity of the results to the violation of the MAR assumption. Conclusions The findings confirmed the importance of undertaking this kind of analysis in end-of-life care studies.

scholarly journals Extending balance assessment for the generalized propensity score under multiple imputation

2020 ◽  
Vol 9 (1) ◽  
Author(s):  
Anna-Simone J. Frank ◽  
David S. Matteson ◽  
Hiroko K. Solvang ◽  
Angela Lupattelli ◽  
Hedvig Nordeng
Keyword(s):  
Propensity Score ◽  
Multiple Imputation ◽  
Simulated Data ◽  
Missing At Random ◽  
Estimation Bias ◽  
Data Sets ◽  
Balance Assessment ◽  
Generalized Propensity Score ◽  
Standardized Mean Difference ◽  
Definition Of

AbstractThis manuscript extends the definition of the Absolute Standardized Mean Difference (ASMD) for binary exposure (M = 2) to cases for M > 2 on multiple imputed data sets. The Maximal Maximized Standardized Difference (MMSD) and the Maximal Averaged Standardized Difference (MASD) were proposed. For different percentages, missing data were introduced in covariates in the simulated data based on the missing at random (MAR) assumption. We then investigate the performance of these two metric definitions using simulated data of full and imputed data sets. The performance of the MASD and the MMSD were validated by relating the balance metrics to estimation bias. The results show that there is an association between the balance metrics and bias. The proposed balance diagnostics seem therefore appropriate to assess balance for the generalized propensity score (GPS) under multiple imputation.

Using the CES-D scale in a large cohort study and dealing with missing data: Application to the French E3N cohort

2011 ◽  
Vol 26 (S2) ◽  
pp. 572-572
Author(s):  
N. Resseguier ◽  
H. Verdoux ◽  
F. Clavel-Chapelon ◽  
X. Paoletti
Keyword(s):  
Sensitivity Analysis ◽  
Missing Data ◽  
Multiple Imputation ◽  
Missing Values ◽  
Large Population ◽  
Missing At Random ◽  
Population Based ◽  
Missing Value ◽  
Perform Sensitivity Analysis ◽  
The Impact

IntroductionThe CES-D scale is commonly used to assess depressive symptoms (DS) in large population-based studies. Missing values in items of the scale may create biases.ObjectivesTo explore reasons for not completing items of the CES-D scale and to perform sensitivity analysis of the prevalence of DS to assess the impact of different missing data hypotheses.Methods71412 women included in the French E3N cohort returned in 2005 a questionnaire containing the CES-D scale. 45% presented at least one missing value in the scale. An interview study was carried out on a random sample of 204 participants to examine the different hypotheses for the missing value mechanism. The prevalence of DS was estimated according to different methods for handling missing values: complete cases analysis, single imputation, multiple imputation under MAR (missing at random) and MNAR (missing not at random) assumptions.ResultsThe interviews showed that participants were not embarrassed to fill in questions about DS. Potential reasons of nonresponse were identified. MAR and MNAR hypotheses remained plausible and were explored.Among complete responders, the prevalence of DS was 26.1%. After multiple imputation under MAR assumption, it was 28.6%, 29.8% and 31.7% among women presenting up to 4, to 10 and to 20 missing values, respectively. The estimates were robust after applying various scenarios of MNAR data for the sensitivity analysis.ConclusionsThe CES-D scale can easily be used to assess DS in large cohorts. Multiple imputation under MAR assumption allows to reliably handle missing values.

Sign in / Sign up

Export Citation Format

Share Document