Predicting feature imputability in the absence of ground truth
This addresses a practical challenge in data imputation for real-world applications where ground truth is unavailable, though it appears incremental as it builds on existing principal component methods.
The paper tackles the problem of evaluating imputation accuracy without ground truth by proposing a principal component-based method to predict feature imputability, establishing a strong linear relationship between principal component loadings and imputability even under extreme missingness.
Data imputation is the most popular method of dealing with missing values, but in most real life applications, large missing data can occur and it is difficult or impossible to evaluate whether data has been imputed accurately (lack of ground truth). This paper addresses these issues by proposing an effective and simple principal component based method for determining whether individual data features can be accurately imputed - feature imputability. In particular, we establish a strong linear relationship between principal component loadings and feature imputability, even in the presence of extreme missingness and lack of ground truth. This work will have important implications in practical data imputation strategies.