The Impact of Missing Values and Single Imputation upon Rasch Analysis Outcomes: A Simulation Study

  • Imputation becomes common practice through availability of easy-to-use algorithms and software. This study aims to determine if different imputation strategies are robust to the extent and type of missingness, local item dependencies (LID), differential item functioning (DIF), and misfit when doing a Rasch analysis. Four samples were simulated and represented a sample with good metric properties, a sample with LID, a sample with DIF, and a sample with LID and DIF. Missing values were generated with increasing proportion and were either missing at random or completely at random. Four imputation techniques were applied before Rasch analysis and deviation of the results and the quality of fit compared. Imputation strategies showed good performance with less than 15% of missingness. The analysis with missing values performed best in recovering statistical estimates. The best strategy, when doing a Rasch analysis, is the analysis with missing values. If for some reason imputation is necessary, we recommend using the expectation-maximization algorithm.

Export metadata

Additional Services

Search Google Scholar
Author:Carolina Saskia Fellinghauer, Birgit Prodinger, Alan Tennant
Parent Title (English):Journal of Applied Measurement
Publisher:JAM Press
Document Type:Article (peer reviewed)
Publication Year:2018
Tag:Differential item functioning (DIF); Imputation techniques; Local item dependencies (LID); Rasch analysis
First Page:1
Last Page:25
Peer reviewed:Ja
faculties / departments:Fakultät für Angewandte Gesundheits- und Sozialwissenschaften
Dewey Decimal Classification:6 Technik, Medizin, angewandte Wissenschaften / 61 Medizin und Gesundheit / 610 Medizin und Gesundheit