Multiple imputation: a primer

Schafer, Joseph L.

doi:10.1177/096228029900800102

Cited by 2,793 publications

(2,002 citation statements)

References 19 publications

Supporting

Mentioning

1,883

Contrasting

Unclassified

Order By: Relevance

“…Sample sizes for analyses before MI ranged from N = 46–93, and G*Power a priori power analyses estimated that with a small effect size (.20), an N = 80–152 was needed for the repeated measures ANOVA, and N = 386 for the MANOVA. Thus, missing data were replaced through multiple imputation, a frequently used process which replaces missing data through imputing, analysing, and pooling missing data (Schafer, 1999). Multiple imputation is a recommended process for handling missing data regardless of the type of missing data (that is, missing at random, missing completely at random, or missing not at random; Schafer, 1999).…”

Section: Methodsmentioning

confidence: 99%

“…Thus, missing data were replaced through multiple imputation, a frequently used process which replaces missing data through imputing, analysing, and pooling missing data (Schafer, 1999). Multiple imputation is a recommended process for handling missing data regardless of the type of missing data (that is, missing at random, missing completely at random, or missing not at random; Schafer, 1999). All analyses and multiple imputation procedures were conducted through IBM SPSS Statistics version 24.…”

Section: Methodsmentioning

confidence: 99%

See 1 more Smart Citation

Treatment of dissociative disorders and reported changes in inpatient and outpatient cost estimates

Myrick

Webermann

Langeland

et al. 2017

European Journal of Psychotraumatology

View full text Add to dashboard Cite

Background: Interpersonal trauma and trauma-related disorders cost society billions of dollars each year. Because of chronic and severe trauma histories, dissociative disorder (DD) patients spend many years in the mental health system, yet there is limited knowledge about the economic burden associated with DDs. Objective: The current study sought to determine how receiving specialized treatment would relate to estimated costs of inpatient and outpatient mental health services. Method: Patients’ and individual therapists’ reports of inpatient hospitalization days and outpatient treatment sessions were converted into US dollars. DD patients and their clinicians reported on use of inpatient and outpatient services four times over 30 months as part of a larger, naturalistic, international DD treatment study. The baseline sample included 292 clinicians and 280 patients; at the 30-month follow-up, 135 clinicians and 111 patients. Missing data were replaced in analyses to maintain adequate statistical power. The substantial attrition rate (>50%) should be considered in interpreting findings. Results: Longitudinal and cross-sectional analyses of cost estimates based on patient reported inpatient hospitalization significantly decreased over time. Longitudinal cost estimates based on clinician-reported outpatient services also significantly decreased over time. Cross-sectional cost estimates based on patient and clinician reported inpatient hospitalization were significantly lower for patients in later stages of treatment compared to those struggling with safety and stabilization. Cross-sectional cost estimates based on clinician-reported outpatient services were significantly lower for patients in later stages of treatment compared to those in early stages. Conclusions: This pattern of longitudinal and cross-sectional reductions in inpatient and outpatient costs, as reported by both patients and therapists, suggests that DD treatment may be associated with reduced inpatient and outpatient costs over time. Although these preliminary results show decreased mental health care utilization and associated estimated costs, it is not clear whether it was treatment that caused these important changes.

show abstract

Section: Methodsmentioning

confidence: 99%

Section: Methodsmentioning

confidence: 99%

Treatment of dissociative disorders and reported changes in inpatient and outpatient cost estimates

Myrick

Webermann

Langeland

et al. 2017

European Journal of Psychotraumatology

View full text Add to dashboard Cite

show abstract

“…Multiple imputation forms a set of complete datasets based on an imputation model, then uses an analytic model to assess intervention effects on each of the completed datasets. The imputation model used to replace the missing data should always be at least as complex as the analytic model used to examine intervention impact (Collins et al, 2001;Graham et al, 2006Graham et al, , 2007Schafer, 1997Schafer, , 1999Schafer and Graham, 2002).…”

Section: Handling Missing Data In Itt Analyses In Multilevel Rfts-mismentioning

confidence: 99%

“…These complete datasets are then analyzed using standard statistical methods, and inferences on such statistics as the odds ratio for GBG versus internal GBG control DISC diagnoses, are made by accounting for two sources of variation: the average standard errors of the odds ratios (within variation) and the variation in these odds ratios across the multiple imputations standard errors (between variation). Confidence intervals can also be formed according to Rubin (1987Rubin ( , 1996 and Schafer (1997Schafer ( , 1999.MI has some advantages over FIML since it can use this additional information to impute values from a large number of observed extra variables that never appear in the final analysis. FIML can also be used with a modest number of extra variables, collapsing over those not used in the final model as we did in Table 6.…”

Section: Handling Missing Data In Itt Analyses In Multilevel Rfts-mismentioning

confidence: 99%

Methods for testing theory and evaluating impact in randomized field trials: Intent-to-treat analyses for integrating the perspectives of person, place, and time

Brown¹,

Wang²,

Kellam³

et al. 2008

Drug and Alcohol Dependence

150

139

View full text Add to dashboard Cite

Randomized field trials provide unique opportunities to examine the effectiveness of an intervention in real world settings and to test and extend both theory of etiology and theory of intervention. These trials are designed not only to test for overall intervention impact but also to examine how impact varies as a function of individual level characteristics, context, and across time. Examination of such variation in impact requires analytical methods that take into account the trial's multiple nested structure and the evolving changes in outcomes over time. The models that we describe here merge multilevel modeling with growth modeling, allowing for variation in impact to be represented through discrete mixtures-growth mixture models-and nonparametric smooth functions-generalized additive mixed models. These methods are part of an emerging class of multilevel growth mixture models, and we illustrate these with models that examine overall impact and variation in impact. In this paper, we define intent-to-treat analyses in group-randomized multilevel field trials and discuss appropriate ways to identify, examine, and test for variation in impact without inflating the Type I error rate. We describe how to make causal inferences more robust to misspecification of covariates * Corresponding author. Tel.: +1 813 974 6672. E-mail address: hbrown@health.usf.edu (C.H. Brown). Conflict of InterestAuthor Muthén is a co-developer of Mplus, which is discussed in this paper. There are no conflicts of interest. in such analyses and how to summarize and present these interactive intervention effects clearly. Practical strategies for reducing model complexity, checking model fit, and handling missing data are discussed using six randomized field trials to show how these methods may be used across trials randomized at different levels. NIH Public Access

show abstract

“…In contrast, the second option can be realised through the direct likelihood approach, which is the likelihood-based way of using only the available information, see [19]. Various other (mostly nonparametric) methods of using only the observed data are discussed in [24], and for single and multiple imputation techniques in [22,24,28,29,30,31]. As mentioned in the introduction, the fourth option becomes necessary in the case of non-ignorability and MNAR.…”

Section: Approaches To the Analysis Of Recurrent Event Data With Dropoutmentioning

confidence: 99%

The impact of dropouts on the analysis of dose‐finding studies with recurrent event data

Akacha

Benda

2010

Statistics in Medicine

View full text Add to dashboard Cite

SUMMARYThis work is motivated by dose-finding studies, where the number of events per subject within a specified study period form the primary outcome. The aim of these studies is to determine the efficacy of a new drug compared to an active control or placebo. In particular, we are interested in identifying the dose-response relationship and the target dose for which the new drug can be shown to be simultaneously safe and as effective as the control.Given an outcome which is pain-related, we expect a considerable number of patients to drop out before the end of the study period. The impact of missingness on the analysis and diverse models for the missingness process must be carefully considered.The recurrent events are modeled as over-dispersed Poisson process data, with dose as a regressor. Additional covariates such as age may be included. Constant and time-varying rate functions are examined. Based on these models the impact of missingness on the precision of the target dose estimation is evaluated. Diverse models for the missingness process are considered, including dependence on covariates and number of events. The performances of five different analysis methods (a complete case analysis; two analyses using different single imputation techniques; a direct likelihood analysis; and an analysis using pattern-mixture models) are assessed via simulation studies. It is shown that the target dose estimation is robust if the same missingness process holds for the target dose group and the active control group. Furthermore, we demonstrate that this robustness is lost as soon as the missingness mechanisms for the active control and the target dose differ. Among the explored missing data handling methods it is shown that the direct-likelihood approach performs best, even when a missing not at random mechanism holds.

show abstract

Multiple imputation: a primer

Abstract: In recent years, multiple imputation has emerged as a convenient and flexible paradigm for analysing data with missing values. Essential features of multiple imputation are reviewed, with answers to frequently asked questions about using the method in practice.

Cited by 2,793 publications

References 19 publications

Treatment of dissociative disorders and reported changes in inpatient and outpatient cost estimates

Treatment of dissociative disorders and reported changes in inpatient and outpatient cost estimates

Methods for testing theory and evaluating impact in randomized field trials: Intent-to-treat analyses for integrating the perspectives of person, place, and time

The impact of dropouts on the analysis of dose‐finding studies with recurrent event data

Contact Info

Product

Resources

About