When understanding missing data, the first step is to identify the mechanism of missingness. According to Peng et al. (2007), the mechanism and pa

When understanding missing data, the first step is to identify the mechanism of missingness. According to Peng et al. (2007), the mechanism and pattern of missing data have a greater impact on research results than the amount of data missing. Both are vital for researchers to consider before selecting an appropriate procedure for handling missing data. Mechanisms that lead to missing data are typically classified as Missing Completely at Random (MCAR), Missing at Random (MAR), and non-ignorable missing. MCAR occurs when missingness depends on neither observed data nor the missing values themselves (Peng et al., 2007). MAR does not imply ignorable bias; instead, it requires proper handling techniques such as maximum likelihood or multiple imputation (Kang, 2013). Non-ignorable missing occurs when missingness is related to unobserved values and cannot be explained by observed data. This situation is more complex and requires modeling of the missing data mechanism (Peng et al., 2007). Common methods for handling missing data include listwise deletion, pairwise deletion, and mean substitution. Listwise deletion is easy to apply and produces unbiased results if the MCAR assumption is satisfied. However, it reduces sample size and statistical power, and produces biased results when MCAR is not met (Kang, 2013). Pairwise deletion preserves more data by using all available pairs, but the drawback is that parameters are estimated from different subsets, which can yield invalid correlation matrices (Kang, 2013). Mean substitution retains sample size, but it underestimates variability, introduces bias, and adds no new information (Kang, 2013). As a guideline, deletion may be acceptable when data are MCAR and the sample is large. However, if missingness exceeds 5–10% or follows a MAR mechanism, estimation methods such as multiple imputation or maximum likelihood are preferred to preserve validity (Kang, 2013)

                                                              References

Kang, H. (2013). The prevention and handling of the missing data. Korean Journal of Anesthesiology, 64(5), 402. https://doi.org/10.4097/kjae.2013.64.5.402

Peng, C. Y. J., Harwell, M., Liou, S. M., & Ehman, L. H. (2007). Advances In Missing Data Methods and Implications for Educational Research. In Real Data Analysis (pp. 31–78). Information Age Publishing. 

.