Methods for Handling Missing Secondary Respondent Data

Abstract

Secondary respondent data are underutilized because researchers avoid using these data in the presence of substantial missing data. We reviewed, critically evaluated, and tested potential solutions to this problem. Five strategies of dealing with missing partner data are reviewed: complete case analysis, inverse probability weighting, correction with a Heckman selection model, maximum likelihood estimation, and multiple imputation. Two approaches were used to evaluate the performance of these methods. First, we used data from the National Survey of Fertility Barriers (N = 1,666) to estimate a model predicting marital quality based on characteristics of women and their husbands. Second, we conducted a simulation based on these data testing the five methods and compared the results to estimates where the true value was known. We found that the maximum likelihood and multiple imputation methods were advantageous because they allow researchers to utilize all of the available information as well as produce less biased and more efficient estimates.

Handling Missing Values in Longitudinal Panel Data With Multiple Imputation

“…A second, related explanation is that imputing whole waves also involves imputing the dependent variable, which has been found to do little to improve the efficiency of the analysis (Allison, ; Young & Johnson, ) and, with a relatively small number of imputed data sets, may even increase the standard error by introducing unnecessary random error into the estimates (von Hippel, ).…”

“…Another strategy for the analysis of missing data in panel studies is found primarily in structural equation modeling (SEM) software that uses ML methods (often referred to as a full information maximum likelihood ) to estimate the covariance structure in the presence of missing values. These methods assume that the missing data are MAR and yield parameter estimates for similarly structured models that are essentially equivalent to those obtained with MI (Graham, ; Young & Johnson, ). SEM requires the wide structure for the data, and within‐wave and whole‐wave missing data are treated in the same way.…”

“…Existing guidelines tend to be directed toward statisticians or focus on types of longitudinal data rarely found in the family literature, such as randomized clinical trials (e.g., Daniels & Hogan, ; Enders, ; Hedeker & Gibbons, ; National Research Council, ) or data sets with few cases but many waves, such as cross‐national time‐series studies (e.g., Honaker & King, ). Methods for handling missing values have been addressed in the family literature (e.g., Acock, ; Johnson & Young, ; Young & Johnson, ), but these resources focus primarily on cross‐sectional data. Although much of what we know about the approaches to handling missing values in cross‐sectional situations applies to longitudinal panel data, panel data have characteristics that complicate the application of techniques such as multiple imputation (MI).…”

Mental health of adolescents: variations by borderline intellectual functioning and disability

“…The outcomes of interest for this analysis were measured at Wave 6, which is where the majority of missing data arose for this analysis. Multiple imputation is one of the methods available to handle missing data, however it has been found that imputing dependent (outcome) variables does little to improve the efficiency of the analysis [35,36]. Since the majority of our missing data was due to missing outcomes, we chose not to use multiple imputation to handle the missing data.…”

Trans‐Congregational Triadic Closure: Churchgoers’ Networks Within and Beyond the Pews

“…For the most part, main conclusions from the multiple imputed models do not change from the Heckman procedure (though specific discrepancies are noted below). We give priority in the main text to the former analyses, but note the importance of considering multiple estimation strategies in the context of missing data and potential selection biases (Young and Johnson ).…”