Estimating and using propensity score in presence of missing background data: an application to assess the impact of childbearing on wellbeing

Mattei, Alessandra

doi:10.1007/s10260-007-0086-0

Cited by 83 publications

(77 citation statements)

References 17 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Although a low proportion of missing data was observed in our study (4.0%), a multiple imputation (i.e., five simulations) was performed to handle missing background data in the estimation and the use of the propensity scores (Supporting Information) …”

Section: Methodsmentioning

confidence: 99%

Long‐term benefits of inhaled corticosteroids in asthma: the propensity score method

Chanoine

Dumas

Benmerad

et al. 2014

Pharmacoepidemiology and Drug

View full text Add to dashboard Cite

show abstract

Section: Methodsmentioning

confidence: 99%

Long‐term benefits of inhaled corticosteroids in asthma: the propensity score method

Chanoine

Dumas

Benmerad

et al. 2014

Pharmacoepidemiology and Drug

View full text Add to dashboard Cite

show abstract

“…It is important to note, that although our study is cross-sectional, these methods have recently been suggested as a means of accounting for missing data values in longitudinal analysis [26]. The basic idea is that separate PSs are computed for subsets of subjects with different missing information patterns.…”

Section: Resolutions Of Problemsmentioning

confidence: 99%

Statistical Challenges in Studying Complex and Changing Families

Bray

Pequegnat

2011

AIDS Behav

View full text Add to dashboard Cite

This paper describes challenges of studying complex and changing families and provides suggestions for methods of analyses to study these family systems. Five common problems that NIMH-funded investigators have encountered in analyzing family based studies are addressed and analytic solutions for addressing these problems are illustrated using data from Family HIV/AIDS projects. The problems discussed in the paper are (1) differences in subgroup responses to interventions; (2) longitudinal changes in family relationships and covariates that vary by group in nested designs; (3) dealing with missing data over time from attrition and planned missing data due to death; (4) dealing with multiple reports from different family members; and (5) developing concordance measures among family members. The use of multi-level and growth curve modeling techniques to address these problems is illustrated with some of the studies.

show abstract

“…Note that an alternative approach would be to define

X_{o b s}

instead as

R X

(which is equivalent to

X^{*}

) and

X_{m i s}

(1 - R) X

. However, for the purposes of this paper, we use the

X_{o b s}

and

X_{m i s}

notation, following the literature on which our theory builds (D'Agostino & Rubin, ; Mattei, ).…”

Section: Introductionmentioning

confidence: 99%

“…We call the second and third assumptions the conditionally independent treatment (CIT) assumption and the conditionally independent outcomes (CIO) assumption, respectively. The CIT assumption is that missing confounder values are conditionally independent of treatment, given the observed confounder values and the missing indicator, whereas the CIO assumption is that missing confounder values are conditionally independent of the potential outcomes (Mattei, ).

\begin{matrix} true \\ right normalCIT : Z & left ⊥ X_{m i s} | C, X_{o b s}, R . \end{matrix}

\begin{matrix} true \\ right normalCIO : Y false(z false) & left ⊥ X_{m i s} | C, X_{o b s}, R normalfor z = 0, 1 . \end{matrix}

Note that in scenarios with partially observed confounders, the mSITA, CIT, and CIO assumptions replace the SITA assumption with respect to identification of the causal estimand.…”

Section: Introductionmentioning

confidence: 99%

Estimating treatment effects with partially observed covariates using outcome regression with missing indicators

et al. 2020

View full text Add to dashboard Cite

Missing data is a common issue in research using observational studies to investigate the effect of treatments on health outcomes. When missingness occurs only in the covariates, a simple approach is to use missing indicators to handle the partially observed covariates. The missing indicator approach has been criticized for giving biased results in outcome regression. However, recent papers have suggested that the missing indicator approach can provide unbiased results in propensity score analysis under certain assumptions. We consider assumptions under which the missing indicator approach can provide valid inferences, namely, (1) no unmeasured confounding within missingness patterns; either (2a) covariate values of patients with missing data were conditionally independent of treatment or (2b) these values were conditionally independent of outcome; and (3) the outcome model is correctly specified: specifically, the true outcome model does not include interactions between missing indicators and fully observed covariates. We prove that, under the assumptions above, the missing indicator approach with outcome regression can provide unbiased estimates of the average treatment effect. We use a simulation study to investigate the extent of bias in estimates of the treatment effect when the assumptions are violated and we illustrate our findings using data from electronic health records. In conclusion, the missing indicator approach can provide valid inferences for outcome regression, but the plausibility of its assumptions must first be considered carefully.

show abstract

Estimating and using propensity score in presence of missing background data: an application to assess the impact of childbearing on wellbeing

Abstract: Ignorability, Propensity score, Missing data, Childbearing, Wellbeing,

Cited by 83 publications

References 17 publications

Long‐term benefits of inhaled corticosteroids in asthma: the propensity score method

Long‐term benefits of inhaled corticosteroids in asthma: the propensity score method

Statistical Challenges in Studying Complex and Changing Families

Estimating treatment effects with partially observed covariates using outcome regression with missing indicators

Contact Info

Product

Resources

About