A Comparison of Joint Model and Fully Conditional Specification Imputation for Multilevel Missing Data

Mistler, Stephen A.; Enders, Craig K.

doi:10.3102/1076998617690869

Cited by 23 publications

(21 citation statements)

References 44 publications

(73 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Therefore as these differences are not largely relevant in the context of our example we do not discuss these differences in more detail. For a detailed discussion of the formal differences between the JM and FCS approaches in the multilevel context, see Carpenter and Kenward (2013), Enders et al (2016) and Mistler (2017) [9,10,56].…”

Section: Discussionmentioning

confidence: 99%

“…Similar to the JM approach by Schafer and Yucel (2002), this approach also uses a joint MLMM for imputing incomplete variables but treats all variables, complete and incomplete, as outcomes in the imputation model [61]. This approach is slightly less restrictive than the JM approach by Schafer and Yucel (2002) as it allows associations between all variables to vary at different levels, and as a result can be congenial with more complicated multilevel analysis models that assume different associations between variables (both complete and incomplete) at different levels [7,56]. We did not include this approach in our study as our substantive analysis model did not assume such associations.…”

Section: Discussionmentioning

confidence: 99%

See 1 more Smart Citation

Evaluation of approaches for multiple imputation of three-level data

Wijesuriya

Moreno‐Betancur

Lee

2020

BMC Med Res Methodol

View full text Add to dashboard Cite

Background: Three-level data arising from repeated measures on individuals who are clustered within larger units are common in health research studies. Missing data are prominent in such longitudinal studies and multiple imputation (MI) is a popular approach for handling missing data. Extensions of joint modelling and fully conditional specification MI approaches based on multilevel models have been developed for imputing three-level data. Alternatively, it is possible to extend single-and two-level MI methods to impute three-level data using dummy indicators and/or by analysing repeated measures in wide format. However, most implementations, evaluations and applications of these approaches focus on the context of incomplete two-level data. It is currently unclear which approach is preferable for imputing three-level data. Methods: In this study, we investigated the performance of various MI methods for imputing three-level incomplete data when the target analysis model is a three-level random effects model with a random intercept for each level. The MI methods were evaluated via simulations and illustrated using empirical data, based on a case study from the Childhood to Adolescence Transition Study, a longitudinal cohort collecting repeated measures on students who were clustered within schools. In our simulations we considered a number of different scenarios covering a range of different missing data mechanisms, missing data proportions and strengths of level-2 and level-3 intra-cluster correlations. Results: We found that all of the approaches considered produced valid inferences about both the regression coefficient corresponding to the exposure of interest and the variance components under the various scenarios within the simulation study. In the case study, all approaches led to similar results. Conclusion: Researchers may use extensions to the single-and two-level approaches, or the three-level approaches, to adequately handle incomplete three-level data. The two-level MI approaches with dummy indicator extension or the MI approaches based on three-level models will be required in certain circumstances such as when there are longitudinal data measured at irregular time intervals. However, the single-and two-level approaches with the DI extension should be used with caution as the DI approach has been shown to produce biased parameter estimates in certain scenarios.

show abstract

Section: Discussionmentioning

confidence: 99%

Section: Discussionmentioning

confidence: 99%

Evaluation of approaches for multiple imputation of three-level data

Wijesuriya

Moreno‐Betancur

Lee

2020

BMC Med Res Methodol

View full text Add to dashboard Cite

show abstract

“…A variety of JM imputation algorithms for longitudinal data and multilevel missing data are also available in the literature. [14][15][16][17][18] The FCS approach is much more flexible but potentially flawed by an incompatible distribution and does not require a joint multivariate distribution, but it specifies an imputation model for each missing variable using its fully conditional density. Fully means that conditioning is usually done on all other variables in the dataset.…”

Section: Introductionmentioning

confidence: 99%

Multiple imputation with sequential penalized regression

Zahid

Heumann

2018

Stat Methods Med Res

View full text Add to dashboard Cite

Missing data is a common issue that can cause problems in estimation and inference in biomedical, epidemiological and social research. Multiple imputation is an increasingly popular approach for handling missing data. In case of a large number of covariates with missing data, existing multiple imputation software packages may not work properly and often produce errors. We propose a multiple imputation algorithm called mispr based on sequential penalized regression models. Each variable with missing values is assumed to have a different distributional form and is imputed with its own imputation model using the ridge penalty. In the case of a large number of predictors with respect to the sample size, the use of a quadratic penalty guarantees unique estimates for the parameters and leads to better predictions than the usual Maximum Likelihood Estimation (MLE), with a good compromise between bias and variance. As a result, the proposed algorithm performs well and provides imputed values that are better even for a large number of covariates with small samples. The results are compared with the existing software packages mice, VIM and Amelia in simulation studies. The missing at random mechanism was the main assumption in the simulation study. The imputation performance of the proposed algorithm is evaluated with mean squared imputation error and mean absolute imputation error. The mean squared error ([Formula: see text]), parameter estimates with their standard errors and confidence intervals are also computed to compare the performance in the regression context. The proposed algorithm is observed to be a good competitor to the existing algorithms, with smaller mean squared imputation error, mean absolute imputation error and mean squared error. The algorithm's performance becomes considerably better than that of the existing algorithms with increasing number of covariates, especially when the number of predictors is close to or even greater than the sample size. Two real-life datasets are also used to examine the performance of the proposed algorithm using simulations.

show abstract

“…Missing data were imputed using substantive model compatible fully conditional specification (SMC-FCS; . Fully conditional specification imputes missing data for variable y using observed values of y and a vector of observed covariates x (Mistler & Enders 2017).…”

Section: Missing Data Imputationmentioning

confidence: 99%

“…Because of the univariate conditional nature of imputation, once a variable y has been sampled any missing data in x are sampled using information from the imputed y. Thus, variables are imputed one at a time, with the filled-in variable from one step serving as a predictor in all subsequent imputation steps (Enders et al 2017).…”

Section: Missing Data Imputationmentioning

confidence: 99%

Bridging the theory to evidence gap: a systematic review and analysis of individual × environment models of child development

Rankin¹

View full text Add to dashboard Cite

A key theoretical principle of developmental psychology is that the response of individuals to developmental experiences can vary. This principle is exemplified in theoretical models of individual × environment interactions, including diathesis stress, differential susceptivity (biological sensitivity to context), and vantage sensitivity.Despite a theoretical underpinning and growing empirical base, there is considerable variability in evidence for these theoretical interactions and the literature has not always delivered well with definitive findings towards the provision of evidence-based interventions tailored to meet the needs of individual children.This thesis has therefore focused on understanding and bridging the gap between theory and evidence for specific individual × environment explanatory models of child behavioural outcomes.It examined the theory and evidence for moderation of parenting effects by child reactivity using a strategy of targeted review and empirical analysis. The thesis aimed to 1) systematically analyse the current literature to identify any patterns in the use of measurement and methods of analysis that might direct future research into individual × environment models. And, 2) conduct analyses using existing data (Longitudinal Study of Australian Children, Family Life Project) to explore effects of different measurements on individual × environment models; systematically examining effects of different measures of parenting, behavioural outcomes, and reactivity. A diagrammatic outline of the thesis is provided in Figure 1.Findings from both the systematic review and empirical analyses revealed that the gap between theory and evidence is extensive. A review of 542 individual × parenting interactions found that 86 were described as statistically reliable. However, within-and between-studies, these reliable interactions inconsistently supported different theoretical models and sporadically varied with measures of parenting, behaviour, and individual characteristics.Two empirical analysis chapters confirmed these inconsistencies. The first empirical study examined the interaction between parent-reported parenting (warmth and harshness) and parentreported temperament (persistence, introversion, reactivity, difficult composite) predicting parent and teacher reported behavioural outcomes (internalizing, externalizing, pro-social) in a longitudinal sample of 1289 Australian children at ages 4 and 6, and across ages 4, 6, and 8. By transparently examining the full matrix of analytical decisions, the results showed that 13 (2.5%) of 512 interactions were statistically reliable and that temperament moderated the effects of parenting on behaviour. However, these reliable interactions did not consistently support a theoretical model iii and varied with measures of parenting, temperament, and behaviour, the age of the child, and the reporter of behaviour.The second empirical chapter extended this analysis by considering four additional measurements and examining an additional data set of children fr...

show abstract

A Comparison of Joint Model and Fully Conditional Specification Imputation for Multilevel Missing Data

Cited by 23 publications

References 44 publications

Evaluation of approaches for multiple imputation of three-level data

Evaluation of approaches for multiple imputation of three-level data

Multiple imputation with sequential penalized regression

Bridging the theory to evidence gap: a systematic review and analysis of individual × environment models of child development

Contact Info

Product

Resources

About