Impact of Missing Values on the Efficiency of Dispersion Control Charts

Abstract

ABSTRACT Accurate estimation of parameters during Phase I analysis is a critical step to ensure the effectiveness of the quality control process. Missing values are often encountered in data collection. Ignoring or inappropriate handling of such data can significantly affect the precision of the parameter estimates and the reliability of the control charts. Therefore, before estimating the process parameters, it is important to impute missing values in control charts to ensure a high‐quality input representation. This study presents a comprehensive investigation into the impact of various missing value handling techniques during phase I analysis on the subsequent performance of Exponentially Weighted Moving Average (EWMA) control charts for process dispersion. The robustness of the chart's monitoring capability is evaluated under three common imputation strategies: mean imputation, median imputation, and k ‐nearest neighbors ( k ‐NN) imputation. Furthermore, the common practice of listwise deletion, whereby cases with missing data are simply eliminated from the analysis, is also examined to provide a comparative baseline for the imputation approaches. Performance is compared for different sample sizes with varying percentages of missing values. Our results show that the k ‐NN imputation outperforms the other methods and provides more accurate imputation to detect the out‐of‐control process.