Modelling non‐ignorable missing‐data mechanisms with item response theory models

Holman, Rebecca; Glas, Cornelis A.W.

doi:10.1111/j.2044-8317.2005.tb00312.x

Cited by 138 publications

(239 citation statements)

References 44 publications

Supporting

Mentioning

231

Contrasting

Unclassified

Order By: Relevance

“…From the simulations, the Mean Absolute Error (MAE) of the discrimination and difficulty parameter estimates for the simulated items was one of the primary outcome variables in this study, as was the case in previous similar work (e.g., Glas, Pimentel, & Lamers, 2015;Holman & Glas, 2005). In addition, because MAE is expressed in the original units of the population parameter, such as difficulty or discrimination, it provides a direct measure of the mean accuracy of the estimates that can be directly interpreted in the context of the parameter.…”

Section: Methodsmentioning

confidence: 64%

“…For this reason, future research should include such interaction effects in the context of IRT (e.g., differential item functioning). Finally, alternative methods for data imputation, including predictive mean matching (Schenker & Taylor, 1996), and the MIRT based approach described by Holman and Glas (2005) should also be considered in future research focused on missing data and IRT.…”

Section: Discussionmentioning

confidence: 99%

“…The presence of missing item responses may be inevitable in many instances, and has been shown to have a deleterious impact on the estimation of IRT parameters if it is not handled properly (Andreis & Ferrari, 2012;Finch, 2008;Holman & Glas, 2005). This is particularly true for MNAR, and to a lesser degree MAR data.…”

mentioning

confidence: 99%

See 2 more Smart Citations

Recursive Partitioning Methods for Data Imputation in the Context of Item Response Theory: A Monte Carlo Simulation

Edwards

Finch

2018

Psicológica Journal

View full text Add to dashboard Cite

Missing data is a common problem faced by psychometricians and measurement professionals. To address this issue, there are a number of techniques that have been proposed to handle missing data regarding Item Response Theory. These methods include several types of data imputation methods -corrected item mean substitution imputation, response function imputation, multiple imputation, and the EM algorithm, as well as approaches that do not rely on the imputation of missing values -treating the item as not presented, coding missing responses as incorrect, or as fractionally correct. Of these methods, even though multiple imputation has demonstrated the best performance in prior research, higher MAE was still present. Given this higher model parameter estimation MAE for even the best performing missing data methods, this simulation study's goal was to explore the performance of a set of potentially promising data imputation methods based on recursive partitioning. Results of this study demonstrated that approaches that combine multivariate imputation by chained equations and recursive partitioning algorithms yield data with relatively low estimation MAE for both item difficulty and item discrimination. Implications of these findings are discussed.

show abstract

Section: Methodsmentioning

confidence: 64%

Section: Discussionmentioning

confidence: 99%

See 1 more Smart Citation

Recursive Partitioning Methods for Data Imputation in the Context of Item Response Theory: A Monte Carlo Simulation

Edwards

Finch

2018

Psicológica Journal

View full text Add to dashboard Cite

show abstract

“…We compared the model by Holman and Glas (2005), which explicitly accounts for a latent missing propensity to a simpler model in which omissions are ignored. We first investigated the appropriateness of the assumptions made in Holman and Glas's model.…”

Section: Discussionmentioning

confidence: 99%

“…This particular approach tries to take nonignorable omissions into account by jointly modeling the distribution of the ability and the missing propensity (Holman & Glas, 2005;O'Muircheartaigh & Moustaki, 1999). Let v index the person, for v = 1, .…”

Section: Theoretical Backgroundmentioning

confidence: 99%

Taking the Missing Propensity Into Account When Estimating Competence Scores

Kohler

Pohl

Carstensen

2014

Educational and Psychological Measurement

View full text Add to dashboard Cite

When competence tests are administered, subjects frequently omit items. These missing responses pose a threat to correctly estimating the proficiency level. Newer model-based approaches aim to take nonignorable missing data processes into account by incorporating a latent missing propensity into the measurement model. Two assumptions are typically made when using these models: (1) The missing propensity is unidimensional and (2) the missing propensity and the ability are bivariate normally distributed. These assumptions may, however, be violated in real data sets and could, thus, pose a threat to the validity of this approach. The present study focuses on modeling competencies in various domains, using data from a school sample ( = 15,396) and an adult sample ( = 7,256) from the National Educational Panel Study. Our interest was to investigate whether violations of unidimensionality and the normal distribution assumption severely affect the performance of the model-based approach in terms of differences in ability estimates. We propose a model with a competence dimension, a unidimensional missing propensity and a distributional assumption more flexible than a multivariate normal. Using this model for ability estimation results in different ability estimates compared with a model ignoring missing responses. Implications for ability estimation in large-scale assessments are discussed.

show abstract

Rank‐based inference with responses missing not at random

Bindele

Adekpedjou

2018

Can J Statistics

View full text Add to dashboard Cite

Missing data have become almost inevitable whenever data are collected. In this paper, interest is given to responses missing not at random in the context of regression modeling. Many of the existing methods for estimating the model parameters lack robustness or efficiency. We propose a robust and efficient approach towards estimating the true regression parameters when some responses in the regression model are missing not at random. Large sample properties of the proposed estimator are established under mild regularity conditions. Monte Carlo simulation experiments are carried out and they showed that the proposed estimator is more efficient than the least squares estimator whenever the model error distribution is heavy tailed, contaminated or when data contain gross outliers. Finally, the method is illustrated using the ACTG protocol 315 data. The Canadian Journal of Statistics 46: 501–528; 2018 © 2018 Statistical Society of Canada

show abstract

Modelling non‐ignorable missing‐data mechanisms with item response theory models

Cited by 138 publications

References 44 publications

Recursive Partitioning Methods for Data Imputation in the Context of Item Response Theory: A Monte Carlo Simulation

Recursive Partitioning Methods for Data Imputation in the Context of Item Response Theory: A Monte Carlo Simulation

Taking the Missing Propensity Into Account When Estimating Competence Scores

Rank‐based inference with responses missing not at random

Contact Info

Product

Resources

About