Comparative Analysis of Imputation Methods for Missing Environmental Data: A Case Study on Ozone Concentrations

Nurliyana  Juhan; Siti Noradiah  Jamaludin; Yong Zulina  Zubairi; Dg Siti Nurisya Sahirah  Ag Isha; Nur Idayu Ah  Khaliludin

doi:10.37934/ard.134.1.6376

Authors

Nurliyana Juhan Preparatory Center for Science and Technology, University Malaysia Sabah (UMS), Jalan UMS, 88400, Kota Kinabalu, Sabah, Malaysia
Siti Noradiah Jamaludin Institute of Mathematical Sciences, University of Malaya, 50603 Kuala Lumpur, Malaysia
Yong Zulina Zubairi Institute for Advanced Studies, University of Malaya, 50603 Kuala Lumpur, Malaysia
Dg Siti Nurisya Sahirah Ag Isha Faculty of Science and Natural Resources, University Malaysia Sabah (UMS), Jalan UMS, 88400, Kota Kinabalu, Sabah, Malaysia
Nur Idayu Ah Khaliludin Tahmidi Centre, Universiti Sains Islam Malaysia, Bandar Baru Nilai 71800, Nilai, Negeri Sembilan, Malaysia

DOI:

https://doi.org/10.37934/ard.134.1.6376

Keywords:

Imputation method, missing data, mean-before-after, ozone concentrations

Abstract

Handling missing values is crucial to environmental data analysis since missing datasets can lead to biassed results. Using Weibull distributions, this study compared six single-imputation methods (mean, median, mean-before-after (MBA), cubic interpolation, linear interpolation, last observation carried forward (LOCF)) for estimating missing ozone concentration data in Petaling Jaya, Selangor. The present study simulated data for sample sizes of 50 and 150 with varying missing value percentages (5%, 10%, 15%, 20%, and 25%). The performance of each imputation method was evaluated using prediction accuracy, root mean square error (RMSE) and mean absolute error (MAE). The findings suggested that the MBA approach outperformed all examined cases, followed by linear interpolation and LOCF. Conversely, cubic interpolation, mean, and median substitution approaches performed poorly, especially as the proportion of missing data increased. This study emphasises the critical role of selecting appropriate imputation methods to enable accurate and trustworthy environmental data analysis. The findings can help researchers select efficient approaches for addressing missing values in air quality datasets, thus boosting the reliability of environmental studies.