High-dimensional imputation for the social sciences: A comparison of state-of-the-art methods

E. Costantini*, K.M. Lang, T. Reeskens, K. Sijtsma

*Corresponding author for this work

Research output: Contribution to journalArticleScientificpeer-review

38 Downloads (Pure)

Abstract

Including a large number of predictors in the imputation model underlying a multiple imputation (MI) procedure is one of the most challenging tasks imputers face. A variety of high-dimensional MI techniques can help, but there has been limited research on their relative performance. In this study, we investigated a wide range of extant high-dimensional MI techniques that can handle a large number of predictors in the imputation models and general missing data patterns. We assessed the relative performance of seven high-dimensional MI methods with a Monte Carlo simulation study and a resampling study based on real survey data. The performance of the methods was defined by the degree to which they facilitate unbiased and confidence-valid estimates of the parameters of complete data analysis models. We found that using lasso penalty or forward selection to select the predictors used in the MI model and using principal component analysis to reduce the dimensionality of auxiliary data produce the best results.
Original languageEnglish
Number of pages52
JournalSociological Methods & Research
DOIs
Publication statusE-pub ahead of print - 2023

Keywords

  • Cart
  • Multiple imputation
  • High-dimensionality
  • Principal components
  • Random forest
  • Regularized regression

Fingerprint

Dive into the research topics of 'High-dimensional imputation for the social sciences: A comparison of state-of-the-art methods'. Together they form a unique fingerprint.

Cite this